Authors: Sai Teja Erukude, Viswa Chaitanya Marella, Suhasnadh Reddy Veluru
Abstract: The rapid rise of photorealistic images produced from Generative Adversarial Networks (GANs) poses a serious challenge for image forensics and industrial systems requiring reliable content authenticity. This paper uses frequency-domain analysis combined with deep learning to solve the problem of distinguishing StyleGAN-generated images from real ones. Specifically, a two-dimensional Discrete Fourier Transform (2D DFT) was applied to transform images into the Fourier domain, where subtle periodic artifacts become detectable. A ResNet50 neural network is trained on these transformed images to differentiate between real and synthetic ones. The experiments demonstrate that the frequency-domain model achieves a 92.8 percent and an AUC of 0.95, significantly outperforming the equivalent model trained on raw spatial-domain images. These results indicate that the GAN-generated images have unique frequency-domain signatures or "fingerprints". The method proposed highlights the industrial potential of combining signal processing techniques and deep learning to enhance digital forensics and strengthen the trustworthiness of industrial AI systems.
Authors: M\'arcus Vin\'icius Lobo Costa, Sherlon Almeida da Silva, B\'arbara Caroline Benato, Leo Sampaio Ferraz Ribeiro, Moacir Antonelli Ponti
Abstract: This paper addresses the challenges in representation learning of 3D shape features by investigating state-of-the-art backbones paired with both contrastive supervised and self-supervised learning objectives. Computer vision methods struggle with recognizing 3D objects from 2D images, often requiring extensive labeled data and relying on Convolutional Neural Networks (CNNs) that may overlook crucial shape relationships. Our work demonstrates that Vision Transformers (ViTs) based architectures, when paired with modern contrastive objectives, achieve promising results in multi-view 3D analysis on our downstream tasks, unifying contrastive and 3D shape understanding pipelines. For example, supervised contrastive losses reached about 90.6% accuracy on ModelNet10. The use of ViTs and contrastive learning, leveraging ViTs' ability to understand overall shapes and contrastive learning's effectiveness, overcomes the need for extensive labeled data and the limitations of CNNs in capturing crucial shape relationships. The success stems from capturing global shape semantics via ViTs and refining local discriminative features through contrastive optimization. Importantly, our approach is empirical, as it is grounded on extensive experimental evaluation to validate the effectiveness of combining ViTs with contrastive objectives for 3D representation learning.
Authors: Martha Teiko Teye, Ori Maoz, Matthias Rottmann
Abstract: We propose FutrTrack, a modular camera-LiDAR multi-object tracking framework that builds on existing 3D detectors by introducing a transformer-based smoother and a fusion-driven tracker. Inspired by query-based tracking frameworks, FutrTrack employs a multimodal two-stage transformer refinement and tracking pipeline. Our fusion tracker integrates bounding boxes with multimodal bird's-eye-view (BEV) fusion features from multiple cameras and LiDAR without the need for an explicit motion model. The tracker assigns and propagates identities across frames, leveraging both geometric and semantic cues for robust re-identification under occlusion and viewpoint changes. Prior to tracking, we refine sequences of bounding boxes with a temporal smoother over a moving window to refine trajectories, reduce jitter, and improve spatial consistency. Evaluated on nuScenes and KITTI, FutrTrack demonstrates that query-based transformer tracking methods benefit significantly from multimodal sensor features compared with previous single-sensor approaches. With an aMOTA of 74.7 on the nuScenes test set, FutrTrack achieves strong performance on 3D MOT benchmarks, reducing identity switches while maintaining competitive accuracy. Our approach provides an efficient framework for improving transformer-based trackers to compete with other neural-network-based methods even with limited data and without pretraining.
Authors: Kushan Choudhury, Shubhrodeep Roy, Ankur Chanda, Shubhajit Biswas, Somenath Kuiry
Abstract: Deep learning models, especially convolutional neural networks, have achieved impressive results in medical image classification. However, these models often produce overconfident predictions, which can undermine their reliability in critical healthcare settings. While traditional label smoothing offers a simple way to reduce such overconfidence, it fails to consider relationships between classes by treating all non-target classes equally. In this study, we explore the use of Online Label Smoothing (OLS), a dynamic approach that adjusts soft labels throughout training based on the model's own prediction patterns. We evaluate OLS on the large-scale RadImageNet dataset using three widely used architectures: ResNet-50, MobileNetV2, and VGG-19. Our results show that OLS consistently improves both Top-1 and Top-5 classification accuracy compared to standard training methods, including hard labels, conventional label smoothing, and teacher-free knowledge distillation. In addition to accuracy gains, OLS leads to more compact and well-separated feature embeddings, indicating improved representation learning. These findings suggest that OLS not only strengthens predictive performance but also enhances calibration, making it a practical and effective solution for developing trustworthy AI systems in the medical imaging domain.
Authors: Neema Jakisa Owor, Joshua Kofi Asamoah, Tanner Wambui Muturi, Anneliese Jakisa Owor, Blessing Agyei Kyem, Andrews Danyo, Yaw Adu-Gyamfi, Armstrong Aboah
Abstract: Fisheye cameras offer an efficient solution for wide-area traffic surveillance by capturing large fields of view from a single vantage point. However, the strong radial distortion and nonuniform resolution inherent in fisheye imagery introduce substantial challenges for standard object detectors, particularly near image boundaries where object appearance is severely degraded. In this work, we present a detection framework designed to operate robustly under these conditions. Our approach employs a simple yet effective pre and post processing pipeline that enhances detection consistency across the image, especially in regions affected by severe distortion. We train several state-of-the-art detection models on the fisheye traffic imagery and combine their outputs through an ensemble strategy to improve overall detection accuracy. Our method achieves an F1 score of0.6366 on the 2025 AI City Challenge Track 4, placing 8thoverall out of 62 teams. These results demonstrate the effectiveness of our framework in addressing issues inherent to fisheye imagery.
Authors: Damian Bowness, Charalambos Poullis
Abstract: When viewing a 3D Gaussian Splatting (3DGS) model from camera positions significantly outside the training data distribution, substantial visual noise commonly occurs. These artifacts result from the lack of training data in these extrapolated regions, leading to uncertain density, color, and geometry predictions from the model. To address this issue, we propose a novel real-time render-aware filtering method. Our approach leverages sensitivity scores derived from intermediate gradients, explicitly targeting instabilities caused by anisotropic orientations rather than isotropic variance. This filtering method directly addresses the core issue of generative uncertainty, allowing 3D reconstruction systems to maintain high visual fidelity even when users freely navigate outside the original training viewpoints. Experimental evaluation demonstrates that our method substantially improves visual quality, realism, and consistency compared to existing Neural Radiance Field (NeRF)-based approaches such as BayesRays. Critically, our filter seamlessly integrates into existing 3DGS rendering pipelines in real-time, unlike methods that require extensive post-hoc retraining or fine-tuning. Code and results at https://damian-bowness.github.io/EV3DGS
Authors: Shengyu Chen, Shihang Feng, Yi Luo, Xiaowei Jia, Youzuo Lin
Abstract: Ultrasound brain imaging remains challenging due to the large difference in sound speed between the skull and brain tissues and the difficulty of coupling large probes to the skull. This work aims to achieve quantitative transcranial ultrasound by reconstructing an accurate speed-of-sound (SoS) map of the brain. Traditional physics-based full-waveform inversion (FWI) is limited by weak signals caused by skull-induced attenuation, mode conversion, and phase aberration, as well as incomplete spatial coverage since full-aperture arrays are clinically impractical. In contrast, purely data-driven methods that learn directly from raw ultrasound data often fail to model the complex nonlinear and nonlocal wave propagation through bone, leading to anatomically plausible but quantitatively biased SoS maps under low signal-to-noise and sparse-aperture conditions. To address these issues, we propose BrainPuzzle, a hybrid two-stage framework that combines physical modeling with machine learning. In the first stage, reverse time migration (time-reversal acoustics) is applied to multi-angle acquisitions to produce migration fragments that preserve structural details even under low SNR. In the second stage, a transformer-based super-resolution encoder-decoder with a graph-based attention unit (GAU) fuses these fragments into a coherent and quantitatively accurate SoS image. A partial-array acquisition strategy using a movable low-count transducer set improves feasibility and coupling, while the hybrid algorithm compensates for the missing aperture. Experiments on two synthetic datasets show that BrainPuzzle achieves superior SoS reconstruction accuracy and image completeness, demonstrating its potential for advancing quantitative ultrasound brain imaging.
Authors: Huichan Seo, Sieun Choi, Minki Hong, Yi Zhou, Junseo Kim, Lukman Ismaila, Naome Etori, Mehul Agarwal, Zhixuan Liu, Jihie Kim, Jean Oh
Abstract: Generative image models produce striking visuals yet often misrepresent culture. Prior work has examined cultural bias mainly in text-to-image (T2I) systems, leaving image-to-image (I2I) editors underexplored. We bridge this gap with a unified evaluation across six countries, an 8-category/36-subcategory schema, and era-aware prompts, auditing both T2I generation and I2I editing under a standardized protocol that yields comparable diagnostics. Using open models with fixed settings, we derive cross-country, cross-era, and cross-category evaluations. Our framework combines standard automatic metrics, a culture-aware retrieval-augmented VQA, and expert human judgments collected from native reviewers. To enable reproducibility, we release the complete image corpus, prompts, and configurations. Our study reveals three findings: (1) under country-agnostic prompts, models default to Global-North, modern-leaning depictions that flatten cross-country distinctions; (2) iterative I2I editing erodes cultural fidelity even when conventional metrics remain flat or improve; and (3) I2I models apply superficial cues (palette shifts, generic props) rather than era-consistent, context-aware changes, often retaining source identity for Global-South targets. These results highlight that culture-sensitive edits remain unreliable in current systems. By releasing standardized data, prompts, and human evaluation protocols, we provide a reproducible, culture-centered benchmark for diagnosing and tracking cultural bias in generative image models.
Authors: Bernd Pfrommer
Abstract: Reconstructing an intensity image from the events of a moving event camera is a challenging task that is typically approached with neural networks deployed on graphics processing units. This paper presents a much simpler, FIlter Based Asynchronous Reconstruction method (FIBAR). First, intensity changes signaled by events are integrated with a temporal digital IIR filter. To reduce reconstruction noise, stale pixels are detected by a novel algorithm that regulates a window of recently updated pixels. Arguing that for a moving camera, the absence of events at a pixel location likely implies a low image gradient, stale pixels are then blurred with a Gaussian filter. In contrast to most existing methods, FIBAR is asynchronous and permits image read-out at an arbitrary time. It runs on a modern laptop CPU at about 42(140) million events/s with (without) spatial filtering enabled. A few simple qualitative experiments are presented that show the difference in image reconstruction between FIBAR and a neural network-based approach (FireNet). FIBAR's reconstruction is noisier than neural network-based methods and suffers from ghost images. However, it is sufficient for certain tasks such as the detection of fiducial markers. Code is available at https://github.com/ros-event-camera/event_image_reconstruction_fibar
URLs: https://github.com/ros-event-camera/event_image_reconstruction_fibar
Authors: Hui Chen, Xinjie Wang, Xianchao Xiu, Wanquan Liu
Abstract: Tensor low-rank representation (TLRR) has demonstrated significant success in image clustering. However, most existing methods rely on fixed transformations and suffer from poor robustness to noise. In this paper, we propose a novel transformed bilateral tensor low-rank representation model called TBTLRR, which introduces a data-adaptive tensor nuclear norm by learning arbitrary unitary transforms, allowing for more effective capture of global correlations. In addition, by leveraging the bilateral structure of latent tensor data, TBTLRR is able to exploit local correlations between image samples and features. Furthermore, TBTLRR integrates the $\ell_{1/2}$-norm and Frobenius norm regularization terms for better dealing with complex noise in real-world scenarios. To solve the proposed nonconvex model, we develop an efficient optimization algorithm inspired by the alternating direction method of multipliers (ADMM) and provide theoretical convergence. Extensive experiments validate its superiority over the state-of-the-art methods in clustering. The code will be available at https://github.com/xianchaoxiu/TBTLRR.
Authors: Lorenzo Arboit, Dennis N. Schneider, Britty Baby, Vinkle Srivastav, Pietro Mascagni, Nicolas Padoy
Abstract: Video-based assessment and surgical data science can advance surgical training, research, and quality improvement. However, widespread use remains limited by heterogeneous recording formats and privacy concerns associated with video sharing. We present Endoshare, a source-available, cross-platform application for merging, standardizing, and de-identifying endoscopic videos in minimally invasive surgery. Development followed the software development life cycle with iterative, user-centered feedback. During the analysis phase, an internal survey of clinicians and computer scientists based on ten usability heuristics identified key requirements that guided a privacy-by-design architecture. In the testing phase, an external clinician survey combined the same heuristics with Technology Acceptance Model constructs to assess usability and adoption, complemented by benchmarking across different hardware configurations. Four clinicians and four computer scientists initially tested the prototype, reporting high usability (4.68 +/- 0.40/5 and 4.03 +/- 0.51/5), with the lowest score (4.00 +/- 0.93/5) relating to label clarity. After refinement, the testing phase surveyed ten surgeons who reported high perceived usefulness (5.07 +/- 1.75/7), ease of use (5.15 +/- 1.71/7), heuristic usability (4.38 +/- 0.48/5), and strong recommendation (9.20 +/- 0.79/10). Processing time varied with processing mode, video duration (both p <= 0.001), and machine computational power (p = 0.041). Endoshare provides a transparent, user-friendly pipeline for standardized, privacy-preserving surgical video management. Compliance certification and broader interoperability validation are needed to establish it as a deployable alternative to proprietary systems. The software is available at https://camma-public.github.io/Endoshare/
Authors: Hao Yu, Haoyu Chen, Yan Jiang, Wei Peng, Zhaodong Sun, Samuel Kaski, Guoying Zhao
Abstract: Self-attention (SA) has become the cornerstone of modern vision backbones for its powerful expressivity over traditional Convolutions (Conv). However, its quadratic complexity remains a critical bottleneck for practical applications. Given that Conv offers linear complexity and strong visual priors, continuing efforts have been made to promote the renaissance of Conv. However, a persistent performance chasm remains, highlighting that these modernizations have not yet captured the intrinsic expressivity that defines SA. In this paper, we re-examine the design of the CNNs, directed by a key question: what principles give SA its edge over Conv? As a result, we reveal two fundamental insights that challenge the long-standing design intuitions in prior research (e.g., Receptive field). The two findings are: (1) \textit{Adaptive routing}: SA dynamically regulates positional information flow according to semantic content, whereas Conv employs static kernels uniformly across all positions. (2) \textit{Lateral inhibition}: SA induces score competition among token weighting, effectively suppressing redundancy and sharpening representations, whereas Conv filters lack such inhibitory dynamics and exhibit considerable redundancy. Based on this, we propose \textit{Attentive Convolution} (ATConv), a principled reformulation of the convolutional operator that intrinsically injects these principles. Interestingly, with only $3\times3$ kernels, ATConv consistently outperforms various SA mechanisms in fundamental vision tasks. Building on ATConv, we introduce AttNet, a CNN family that can attain \textbf{84.4\%} ImageNet-1K Top-1 accuracy with only 27M parameters. In diffusion-based image generation, replacing all SA with the proposed $3\times 3$ ATConv in SiT-XL/2 reduces ImageNet FID by 0.15 in 400k steps with faster sampling. Code is available at: github.com/price112/Attentive-Convolution.
Authors: Jiho Park, Sieun Choi, Jaeyoon Seo, Jihie Kim
Abstract: Although recent advancements in diffusion models have significantly enriched the quality of generated images, challenges remain in synthesizing pixel-based human-drawn sketches, a representative example of abstract expression. To combat these challenges, we propose StableSketcher, a novel framework that empowers diffusion models to generate hand-drawn sketches with high prompt fidelity. Within this framework, we fine-tune the variational autoencoder to optimize latent decoding, enabling it to better capture the characteristics of sketches. In parallel, we integrate a new reward function for reinforcement learning based on visual question answering, which improves text-image alignment and semantic consistency. Extensive experiments demonstrate that StableSketcher generates sketches with improved stylistic fidelity, achieving better alignment with prompts compared to the Stable Diffusion baseline. Additionally, we introduce SketchDUO, to the best of our knowledge, the first dataset comprising instance-level sketches paired with captions and question-answer pairs, thereby addressing the limitations of existing datasets that rely on image-label pairs. Our code and dataset will be made publicly available upon acceptance.
Authors: Ziheng Zhang, Xinyue Ma, Arpita Chowdhury, Elizabeth G. Campolongo, Matthew J. Thompson, Net Zhang, Samuel Stevens, Hilmar Lapp, Tanya Berger-Wolf, Yu Su, Wei-Lun Chao, Jianyang Gu
Abstract: This work investigates descriptive captions as an additional source of supervision for biological multimodal foundation models. Images and captions can be viewed as complementary samples from the latent morphospace of a species, each capturing certain biological traits. Incorporating captions during training encourages alignment with this shared latent structure, emphasizing potentially diagnostic characters while suppressing spurious correlations. The main challenge, however, lies in obtaining faithful, instance-specific captions at scale. This requirement has limited the utilization of natural language supervision in organismal biology compared with many other scientific domains. We complement this gap by generating synthetic captions with multimodal large language models (MLLMs), guided by Wikipedia-derived visual information and taxon-tailored format examples. These domain-specific contexts help reduce hallucination and yield accurate, instance-based descriptive captions. Using these captions, we train BIOCAP (i.e., BIOCLIP with Captions), a biological foundation model that captures rich semantics and achieves strong performance in species classification and text-image retrieval. These results demonstrate the value of descriptive captions beyond labels in bridging biological images with multimodal foundation models.
Authors: Prithvi Raj Singh, Raju Gottumukkala, Anthony S. Maida, Alan B. Barhorst, Vijaya Gopu
Abstract: While computer vision has advanced considerably for general object detection and tracking, the specific problem of fast-moving tiny objects remains underexplored. This paper addresses the significant challenge of detecting and tracking rapidly moving small objects using an RGB-D camera. Our novel system combines deep learning-based detection with physics-based tracking to overcome the limitations of existing approaches. Our contributions include: (1) a comprehensive system design for object detection and tracking of fast-moving small objects in 3D space, (2) an innovative physics-based tracking algorithm that integrates kinematics motion equations to handle outliers and missed detections, and (3) an outlier detection and correction module that significantly improves tracking performance in challenging scenarios such as occlusions and rapid direction changes. We evaluated our proposed system on a custom racquetball dataset. Our evaluation shows our system surpassing kalman filter based trackers with up to 70\% less Average Displacement Error. Our system has significant applications for improving robot perception on autonomous platforms and demonstrates the effectiveness of combining physics-based models with deep learning approaches for real-time 3D detection and tracking of challenging small objects.
Authors: Hyunjun Jung, Hae-Gon Jeon
Abstract: A concept of light-fields computed from multiple view images on regular grids has proven its benefit for scene representations, and supported realistic renderings of novel views and photographic effects such as refocusing and shallow depth of field. In spite of its effectiveness of light flow computations, obtaining light fields requires either computational costs or specialized devices like a bulky camera setup and a specialized microlens array. In an effort to broaden its benefit and applicability, in this paper, we propose a novel view synthesis method for light field generation from only single images, named inverse image-based rendering. Unlike previous attempts to implicitly rebuild 3D geometry or to explicitly represent objective scenes, our method reconstructs light flows in a space from image pixels, which behaves in the opposite way to image-based rendering. To accomplish this, we design a neural rendering pipeline to render a target ray in an arbitrary viewpoint. Our neural renderer first stores the light flow of source rays from the input image, then computes the relationships among them through cross-attention, and finally predicts the color of the target ray based on these relationships. After the rendering pipeline generates the first novel view from a single input image, the generated out-of-view contents are updated to the set of source rays. This procedure is iteratively performed while ensuring the consistent generation of occluded contents. We demonstrate that our inverse image-based rendering works well with various challenging datasets without any retraining or finetuning after once trained on synthetic dataset, and outperforms relevant state-of-the-art novel view synthesis methods.
Authors: Jiachen Liang, Ruibing Hou, Minyang Hu, Hong Chang, Shiguang Shan, Xilin Chen
Abstract: Out-of-distribution (OOD) detection is critical for ensuring the reliability of deep learning models in open-world applications. While post-hoc methods are favored for their efficiency and ease of deployment, existing approaches often underexploit the rich information embedded in the model's logits space. In this paper, we propose LogitGap, a novel post-hoc OOD detection method that explicitly exploits the relationship between the maximum logit and the remaining logits to enhance the separability between in-distribution (ID) and OOD samples. To further improve its effectiveness, we refine LogitGap by focusing on a more compact and informative subset of the logit space. Specifically, we introduce a training-free strategy that automatically identifies the most informative logits for scoring. We provide both theoretical analysis and empirical evidence to validate the effectiveness of our approach. Extensive experiments on both vision-language and vision-only models demonstrate that LogitGap consistently achieves state-of-the-art performance across diverse OOD detection scenarios and benchmarks. Code is available at https://github.com/GIT-LJc/LogitGap.
Authors: Penghao Wang, Yiyang He, Xin Lv, Yukai Zhou, Lan Xu, Jingyi Yu, Jiayuan Gu
Abstract: Understanding objects at the level of their constituent parts is fundamental to advancing computer vision, graphics, and robotics. While datasets like PartNet have driven progress in 3D part understanding, their reliance on untextured geometries and expert-dependent annotation limits scalability and usability. We introduce PartNeXt, a next-generation dataset addressing these gaps with over 23,000 high-quality, textured 3D models annotated with fine-grained, hierarchical part labels across 50 categories. We benchmark PartNeXt on two tasks: (1) class-agnostic part segmentation, where state-of-the-art methods (e.g., PartField, SAMPart3D) struggle with fine-grained and leaf-level parts, and (2) 3D part-centric question answering, a new benchmark for 3D-LLMs that reveals significant gaps in open-vocabulary part grounding. Additionally, training Point-SAM on PartNeXt yields substantial gains over PartNet, underscoring the dataset's superior quality and diversity. By combining scalable annotation, texture-aware labels, and multi-task evaluation, PartNeXt opens new avenues for research in structured 3D understanding.
Authors: Eduardo R. Corral-Soto, Yang Liu, Yuan Ren, Bai Dongfeng, Liu Bingbing
Abstract: In Autonomous Driving, cyclists belong to the safety-critical class of Vulnerable Road Users (VRU), and accurate estimation of their pose is critical for cyclist crossing intention classification, behavior prediction, and collision avoidance. Unlike rigid objects, articulated bicycles are composed of movable rigid parts linked by joints and constrained by a kinematic structure. 6D pose methods can estimate the 3D rotation and translation of rigid bicycles, but 6D becomes insufficient when the steering/pedals angles of the bicycle vary. That is because: 1) varying the articulated pose of the bicycle causes its 3D bounding box to vary as well, and 2) the 3D box orientation is not necessarily aligned to the orientation of the steering which determines the actual intended travel direction. In this work, we introduce a method for category-level 8D pose estimation for articulated bicycles and cyclists from a single RGB image. Besides being able to estimate the 3D translation and rotation of a bicycle from a single image, our method also estimates the rotations of its steering handles and pedals with respect to the bicycle body frame. These two new parameters enable the estimation of a more fine-grained bicycle pose state and travel direction. Our proposed model jointly estimates the 8D pose and the 3D Keypoints of articulated bicycles, and trains with a mix of synthetic and real image data to generalize on real images. We include an evaluation section where we evaluate the accuracy of our estimated 8D pose parameters, and our method shows promising results by achieving competitive scores when compared against state-of-the-art category-level 6D pose estimators that use rigid canonical object templates for matching.
Authors: Xudong Yan, Songhe Feng
Abstract: Compositional Zero-Shot Learning (CZSL) aims to recognize novel attribute-object compositions based on the knowledge learned from seen ones. Existing methods suffer from performance degradation caused by the distribution shift of label space at test time, which stems from the inclusion of unseen compositions recombined from attributes and objects. To overcome the challenge, we propose a novel approach that accumulates comprehensive knowledge in both textual and visual modalities from unsupervised data to update multimodal prototypes at test time. Building on this, we further design an adaptive update weight to control the degree of prototype adjustment, enabling the model to flexibly adapt to distribution shift during testing. Moreover, a dynamic priority queue is introduced that stores high-confidence images to acquire visual knowledge from historical images for inference. Considering the semantic consistency of multimodal knowledge, we align textual and visual prototypes by multimodal collaborative representation learning. Extensive experiments indicate that our approach achieves state-of-the-art performance on four benchmark datasets under both closed-world and open-world settings. Code will be available at https://github.com/xud-yan/TOMCAT .
Authors: Insu Jeon, Wonkwang Lee, Myeongjang Pyeon, Gunhee Kim
Abstract: We propose a new GAN-based unsupervised model for disentangled representation learning. The new model is discovered in an attempt to utilize the Information Bottleneck (IB) framework to the optimization of GAN, thereby named IB-GAN. The architecture of IB-GAN is partially similar to that of InfoGAN but has a critical difference; an intermediate layer of the generator is leveraged to constrain the mutual information between the input and the generated output. The intermediate stochastic layer can serve as a learnable latent distribution that is trained with the generator jointly in an end-to-end fashion. As a result, the generator of IB-GAN can harness the latent space in a disentangled and interpretable manner. With the experiments on dSprites and Color-dSprites dataset, we demonstrate that IB-GAN achieves competitive disentanglement scores to those of state-of-the-art \b{eta}-VAEs and outperforms InfoGAN. Moreover, the visual quality and the diversity of samples generated by IB-GAN are often better than those by \b{eta}-VAEs and Info-GAN in terms of FID score on CelebA and 3D Chairs dataset.
Authors: Yun Wang, Junjie Hu, Qiaole Dong, Yongjian Zhang, Yanwei Fu, Tin Lun Lam, Dapeng Wu
Abstract: Temporally consistent depth estimation from stereo video is critical for real-world applications such as augmented reality, where inconsistent depth estimation disrupts the immersion of users. Despite its importance, this task remains challenging due to the difficulty in modeling long-term temporal consistency in a computationally efficient manner. Previous methods attempt to address this by aggregating spatio-temporal information but face a fundamental trade-off: limited temporal modeling provides only modest gains, whereas capturing long-range dependencies significantly increases computational cost. To address this limitation, we introduce a memory buffer for modeling long-range spatio-temporal consistency while achieving efficient dynamic stereo matching. Inspired by the two-stage decision-making process in humans, we propose a \textbf{P}ick-and-\textbf{P}lay \textbf{M}emory (PPM) construction module for dynamic \textbf{Stereo} matching, dubbed as \textbf{PPMStereo}. PPM consists of a `pick' process that identifies the most relevant frames and a `play' process that weights the selected frames adaptively for spatio-temporal aggregation. This two-stage collaborative process maintains a compact yet highly informative memory buffer while achieving temporally consistent information aggregation. Extensive experiments validate the effectiveness of PPMStereo, demonstrating state-of-the-art performance in both accuracy and temporal consistency. % Notably, PPMStereo achieves 0.62/1.11 TEPE on the Sintel clean/final (17.3\% \& 9.02\% improvements over BiDAStereo) with fewer computational costs. Codes are available at \textcolor{blue}{https://github.com/cocowy1/PPMStereo}.
Authors: Aaron Appelle, Jerome P. Lynch
Abstract: Large-scale video generation models have demonstrated high visual realism in diverse contexts, spurring interest in their potential as general-purpose world simulators. Existing benchmarks focus on individual subjects rather than scenes with multiple interacting people. However, the plausibility of multi-agent dynamics in generated videos remains unverified. We propose a rigorous evaluation protocol to benchmark text-to-video (T2V) and image-to-video (I2V) models as implicit simulators of pedestrian dynamics. For I2V, we leverage start frames from established datasets to enable comparison with a ground truth video dataset. For T2V, we develop a prompt suite to explore diverse pedestrian densities and interactions. A key component is a method to reconstruct 2D bird's-eye view trajectories from pixel-space without known camera parameters. Our analysis reveals that leading models have learned surprisingly effective priors for plausible multi-agent behavior. However, failure modes like merging and disappearing people highlight areas for future improvement.
Authors: Xinyi Hu, Yuran Wang, Yue Li, Wenxuan Liu, Zheng Wang
Abstract: Temporal Intention Localization (TIL) is crucial for video surveillance, focusing on identifying varying levels of suspicious intentions to improve security monitoring. However, existing discrete classification methods fail to capture the continuous nature of suspicious intentions, limiting early intervention and explainability. In this paper, we propose the Suspicion Progression Analysis Network (SPAN), which shifts from discrete classification to continuous regression, enabling the capture of fluctuating and evolving suspicious intentions. We reveal that suspicion exhibits long-term dependencies and cumulative effects, similar to Temporal Point Process (TPP) theory. Based on these insights, we define a suspicion score formula that models continuous changes while accounting for temporal characteristics. We also introduce Suspicion Coefficient Modulation, which adjusts suspicion coefficients using multimodal information to reflect the varying impacts of suspicious actions. Additionally, the Concept-Anchored Mapping method is proposed to link suspicious actions to predefined intention concepts, offering insights into both the actions and their potential underlying intentions. Extensive experiments on the HAI dataset show that SPAN significantly outperforms existing methods, reducing MSE by 19.8% and improving average mAP by 1.78%. Notably, SPAN achieves a 2.74% mAP gain in low-frequency cases, demonstrating its superior ability to capture subtle behavioral changes. Compared to discrete classification systems, our continuous suspicion modeling approach enables earlier detection and proactive intervention, greatly enhancing system explainability and practical utility in security applications.
Authors: Minh Sao Khue Luu, Margaret V. Benedichuk, Ekaterina I. Roppert, Roman M. Kenzhin, Bair N. Tuchinov
Abstract: The development of foundation models for brain MRI depends critically on the scale, diversity, and consistency of available data, yet systematic assessments of these factors remain scarce. In this study, we analyze 54 publicly accessible brain MRI datasets encompassing over 538,031 to provide a structured, multi-level overview tailored to foundation model development. At the dataset level, we characterize modality composition, disease coverage, and dataset scale, revealing strong imbalances between large healthy cohorts and smaller clinical populations. At the image level, we quantify voxel spacing, orientation, and intensity distributions across 15 representative datasets, demonstrating substantial heterogeneity that can influence representation learning. We then perform a quantitative evaluation of preprocessing variability, examining how intensity normalization, bias field correction, skull stripping, spatial registration, and interpolation alter voxel statistics and geometry. While these steps improve within-dataset consistency, residual differences persist between datasets. Finally, feature-space case study using a 3D DenseNet121 shows measurable residual covariate shift after standardized preprocessing, confirming that harmonization alone cannot eliminate inter-dataset bias. Together, these analyses provide a unified characterization of variability in public brain MRI resources and emphasize the need for preprocessing-aware and domain-adaptive strategies in the design of generalizable brain MRI foundation models.
Authors: Bingjie Gao, Qianli Ma, Xiaoxue Wu, Shuai Yang, Guanzhou Lan, Haonan Zhao, Jiaxuan Chen, Qingyang Liu, Yu Qiao, Xinyuan Chen, Yaohui Wang, Li Niu
Abstract: Prompt design plays a crucial role in text-to-video (T2V) generation, yet user-provided prompts are often short, unstructured, and misaligned with training data, limiting the generative potential of diffusion-based T2V models. We present \textbf{RAPO++}, a cross-stage prompt optimization framework that unifies training-data--aligned refinement, test-time iterative scaling, and large language model (LLM) fine-tuning to substantially improve T2V generation without modifying the underlying generative backbone. In \textbf{Stage 1}, Retrieval-Augmented Prompt Optimization (RAPO) enriches user prompts with semantically relevant modifiers retrieved from a relation graph and refactors them to match training distributions, enhancing compositionality and multi-object fidelity. \textbf{Stage 2} introduces Sample-Specific Prompt Optimization (SSPO), a closed-loop mechanism that iteratively refines prompts using multi-source feedback -- including semantic alignment, spatial fidelity, temporal coherence, and task-specific signals such as optical flow -- yielding progressively improved video generation quality. \textbf{Stage 3} leverages optimized prompt pairs from SSPO to fine-tune the rewriter LLM, internalizing task-specific optimization patterns and enabling efficient, high-quality prompt generation even before inference. Extensive experiments across five state-of-the-art T2V models and five benchmarks demonstrate that RAPO++ achieves significant gains in semantic alignment, compositional reasoning, temporal stability, and physical plausibility, outperforming existing methods by large margins. Our results highlight RAPO++ as a model-agnostic, cost-efficient, and scalable solution that sets a new standard for prompt optimization in T2V generation. The code is available at https://github.com/Vchitect/RAPO.
Authors: Yanghao Wang, Zhen Wang, Long Chen
Abstract: Recent advances in pre-trained text-to-image flow models have enabled remarkable progress in text-based image editing. Mainstream approaches always adopt a corruption-then-restoration paradigm, where the source image is first corrupted into an ``intermediate state'' and then restored to the target image under the prompt guidance. However, current methods construct this intermediate state in a target-agnostic manner, i.e., they primarily focus on realizing source image reconstruction while neglecting the semantic gaps towards the specific editing target. This design inherently results in limited editability or inconsistency when the desired modifications substantially deviate from the source. In this paper, we argue that the intermediate state should be target-aware, i.e., selectively corrupting editing-relevant contents while preserving editing-irrelevant ones. To this end, we propose FlowCycle, a novel inversion-free and flow-based editing framework that parameterizes corruption with learnable noises and optimizes them through a cycle-consistent process. By iteratively editing the source to the target and recovering back to the source with dual consistency constraints, FlowCycle learns to produce a target-aware intermediate state, enabling faithful modifications while preserving source consistency. Extensive ablations have demonstrated that FlowCycle achieves superior editing quality and consistency over state-of-the-art methods.
Authors: Talha Ilyas, Duong Nhu, Allison Thomas, Arie Levin, Lim Wei Yap, Shu Gong, David Vera Anaya, Yiwen Jiang, Deval Mehta, Ritesh Warty, Vinayak Smith, Maya Reddy, Euan Wallace, Wenlong Cheng, Zongyuan Ge, Faezeh Marzbanrad
Abstract: Accurate fetal movement (FM) detection is essential for assessing prenatal health, as abnormal movement patterns can indicate underlying complications such as placental dysfunction or fetal distress. Traditional methods, including maternal perception and cardiotocography (CTG), suffer from subjectivity and limited accuracy. To address these challenges, we propose Contrastive Ultrasound Video Representation Learning (CURL), a novel self-supervised learning framework for FM detection from extended fetal ultrasound video recordings. Our approach leverages a dual-contrastive loss, incorporating both spatial and temporal contrastive learning, to learn robust motion representations. Additionally, we introduce a task-specific sampling strategy, ensuring the effective separation of movement and non-movement segments during self-supervised training, while enabling flexible inference on arbitrarily long ultrasound recordings through a probabilistic fine-tuning approach. Evaluated on an in-house dataset of 92 subjects, each with 30-minute ultrasound sessions, CURL achieves a sensitivity of 78.01% and an AUROC of 81.60%, demonstrating its potential for reliable and objective FM analysis. These results highlight the potential of self-supervised contrastive learning for fetal movement analysis, paving the way for improved prenatal monitoring and clinical decision-making.
Authors: Jiahuan Wang, Yuxin Chen, Jun Yu, Guangming Lu, Wenjie Pei
Abstract: Adapting pretrained diffusion-based generative models for text-driven image editing with negligible tuning overhead has demonstrated remarkable potential. A classical adaptation paradigm, as followed by these methods, first infers the generative trajectory inversely for a given source image by image inversion, then performs image editing along the inferred trajectory guided by the target text prompts. However, the performance of image editing is heavily limited by the approximation errors introduced during image inversion by diffusion models, which arise from the absence of exact supervision in the intermediate generative steps. To circumvent this issue, we investigate the parameter-efficient adaptation of VQ-based generative models for image editing, and leverage their inherent characteristic that the exact intermediate quantized representations of a source image are attainable, enabling more effective supervision for precise image inversion. Specifically, we propose \emph{EditInfinity}, which adapts \emph{Infinity}, a binary-quantized generative model, for image editing. We propose an efficient yet effective image inversion mechanism that integrates text prompting rectification and image style preservation, enabling precise image inversion. Furthermore, we devise a holistic smoothing strategy which allows our \emph{EditInfinity} to perform image editing with high fidelity to source images and precise semantic alignment to the text prompts. Extensive experiments on the PIE-Bench benchmark across "add", "change", and "delete" editing operations, demonstrate the superior performance of our model compared to state-of-the-art diffusion-based baselines. Code available at: https://github.com/yx-chen-ust/EditInfinity.
Authors: Ge Zheng, Jiaye Qian, Jiajin Tang, Sibei Yang
Abstract: Large Vision-Language Models (LVLMs) have made significant progress in recent years but are also prone to hallucination issues. They exhibit more hallucinations in longer, free-form responses, often attributed to accumulated uncertainties. In this paper, we ask: Does increased hallucination result solely from length-induced errors, or is there a deeper underlying mechanism? After a series of preliminary experiments and findings, we suggest that the risk of hallucinations is not caused by length itself but by the increased reliance on context for coherence and completeness in longer responses. Building on these insights, we propose a novel "induce-detect-suppress" framework that actively induces hallucinations through deliberately designed contexts, leverages induced instances for early detection of high-risk cases, and ultimately suppresses potential object-level hallucinations during actual decoding. Our approach achieves consistent, significant improvements across all benchmarks, demonstrating its efficacy. The strong detection and improved hallucination mitigation not only validate our framework but, more importantly, re-validate our hypothesis on context. Rather than solely pursuing performance gains, this study aims to provide new insights and serves as a first step toward a deeper exploration of hallucinations in LVLMs' longer responses.
Authors: Runsong Zhu, Ka-Hei Hui, Zhengzhe Liu, Qianyi Wu, Weiliang Tang, Shi Qiu, Pheng-Ann Heng, Chi-Wing Fu
Abstract: Open-vocabulary 3D segmentation is a fundamental yet challenging task, requiring a mutual understanding of both segmentation and language. However, existing Gaussian-splatting-based methods rely either on a single 3D language field, leading to inferior segmentation, or on pre-computed class-agnostic segmentations, suffering from error accumulation. To address these limitations, we present COS3D, a new collaborative prompt-segmentation framework that contributes to effectively integrating complementary language and segmentation cues throughout its entire pipeline. We first introduce the new concept of collaborative field, comprising an instance field and a language field, as the cornerstone for collaboration. During training, to effectively construct the collaborative field, our key idea is to capture the intrinsic relationship between the instance field and language field, through a novel instance-to-language feature mapping and designing an efficient two-stage training strategy. During inference, to bridge distinct characteristics of the two fields, we further design an adaptive language-to-instance prompt refinement, promoting high-quality prompt-segmentation inference. Extensive experiments not only demonstrate COS3D's leading performance over existing methods on two widely-used benchmarks but also show its high potential to various applications,~\ie, novel image-based 3D segmentation, hierarchical segmentation, and robotics. The code is publicly available at \href{https://github.com/Runsong123/COS3D}{https://github.com/Runsong123/COS3D}.
URLs: https://github.com/Runsong123/COS3D, https://github.com/Runsong123/COS3D
Authors: Minseok Kang, Minhyeok Lee, Minjung Kim, Donghyeong Kim, Sangyoun Lee
Abstract: Video Temporal Grounding (VTG) aims to localize temporal segments in long, untrimmed videos that align with a given natural language query. This task typically comprises two subtasks: Moment Retrieval (MR) and Highlight Detection (HD). While recent advances have been progressed by powerful pretrained vision-language models such as CLIP and InternVideo2, existing approaches commonly treat all text tokens uniformly during crossmodal attention, disregarding their distinct semantic roles. To validate the limitations of this approach, we conduct controlled experiments demonstrating that VTG models overly rely on [EOS]-driven global semantics while failing to effectively utilize word-level signals, which limits their ability to achieve fine-grained temporal alignment. Motivated by this limitation, we propose DualGround, a dual-branch architecture that explicitly separates global and local semantics by routing the [EOS] token through a sentence-level path and clustering word tokens into phrase-level units for localized grounding. Our method introduces (1) tokenrole- aware cross modal interaction strategies that align video features with sentence-level and phrase-level semantics in a structurally disentangled manner, and (2) a joint modeling framework that not only improves global sentence-level alignment but also enhances finegrained temporal grounding by leveraging structured phrase-aware context. This design allows the model to capture both coarse and localized semantics, enabling more expressive and context-aware video grounding. DualGround achieves state-of-the-art performance on both Moment Retrieval and Highlight Detection tasks across QVHighlights and Charades- STA benchmarks, demonstrating the effectiveness of disentangled semantic modeling in video-language alignment.
Authors: Shuhan Hu, Yiru Li, Yuanyuan Li, Yingying Zhu
Abstract: Cross-view object geo-localization enables high-precision object localization through cross-view matching, with critical applications in autonomous driving, urban management, and disaster response. However, existing methods rely on keypoint-based positional encoding, which captures only 2D coordinates while neglecting object shape information, resulting in sensitivity to annotation shifts and limited cross-view matching capability. To address these limitations, we propose a mask-based positional encoding scheme that leverages segmentation masks to capture both spatial coordinates and object silhouettes, thereby upgrading the model from "location-aware" to "object-aware." Furthermore, to tackle the challenge of large-span objects (e.g., elongated buildings) in satellite imagery, we design a context enhancement module. This module employs horizontal and vertical strip convolutional kernels to extract long-range contextual features, enhancing feature discrimination among strip-like objects. Integrating MPE and CEM, we present EDGeo, an end-to-end framework for robust cross-view object geo-localization. Extensive experiments on two public datasets (CVOGL and VIGOR-Building) demonstrate that our method achieves state-of-the-art performance, with a 3.39% improvement in localization accuracy under challenging ground-to-satellite scenarios. This work provides a robust positional encoding paradigm and a contextual modeling framework for advancing cross-view geo-localization research.
Authors: Guowei Zhong, Junjie Li, Huaiyu Zhu, Ruohong Huan, Yun Pan
Abstract: In recent years, Multimodal Emotion Recognition (MER) has made substantial progress. Nevertheless, most existing approaches neglect the semantic inconsistencies that may arise across modalities, such as conflicting emotional cues between text and visual inputs. Besides, current methods are often dominated by the text modality due to its strong representational capacity, which can compromise recognition accuracy. To address these challenges, we propose a model termed Calibrated Multimodal Consensus (CMC). CMC introduces a Pseudo Label Generation Module (PLGM) to produce pseudo unimodal labels, enabling unimodal pretraining in a self-supervised fashion. It then employs a Parameter-free Fusion Module (PFM) and a Multimodal Consensus Router (MCR) for multimodal finetuning, thereby mitigating text dominance and guiding the fusion process toward a more reliable consensus. Experimental results demonstrate that CMC achieves performance on par with or superior to state-of-the-art methods across four datasets, CH-SIMS, CH-SIMS v2, CMU-MOSI, and CMU-MOSEI, and exhibits notable advantages in scenarios with semantic inconsistencies on CH-SIMS and CH-SIMS v2. The implementation of this work is publicly accessible at https://github.com/gw-zhong/CMC.
Authors: Saraf Anzum Shreya, MD. Abu Ismail Siddique, Sharaf Tasnim
Abstract: Technologies like smartphones have become an essential in our daily lives. It has made accessible to everyone including visually impaired individuals. With the use of smartphone cameras, image capturing and processing have become more convenient. With the use of smartphones and machine learning, the life of visually impaired can be made a little easier. Daily tasks such as handling money without relying on someone can be troublesome for them. For that purpose this paper presents a real-time currency detection system designed to assist visually impaired individuals. The proposed model is trained on a dataset containing 30 classes of notes and coins, representing 3 types of currency: US dollar (USD), Euro (EUR), and Bangladeshi taka (BDT). Our approach uses a YOLOv8 nano model with a custom detection head featuring deep convolutional layers and Squeeze-and-Excitation blocks to enhance feature extraction and detection accuracy. Our model has achieved a higher accuracy of 97.73%, recall of 95.23%, f1-score of 95.85% and a mean Average Precision at IoU=0.5 (mAP50(B)) of 97.21\%. Using the voice feedback after the detection would help the visually impaired to identify the currency. This paper aims to create a practical and efficient currency detection system to empower visually impaired individuals independent in handling money.
Authors: Guangyu Dai, Dong Chen, Siliang Tang, Yueting Zhuang
Abstract: Video anomaly detection (VAD) is a challenging task that detects anomalous frames in continuous surveillance videos. Most previous work utilizes the spatio-temporal correlation of visual features to distinguish whether there are abnormalities in video snippets. Recently, some works attempt to introduce multi-modal information, like text feature, to enhance the results of video anomaly detection. However, these works merely incorporate text features into video snippets in a coarse manner, overlooking the significant amount of redundant information that may exist within the video snippets. Therefore, we propose to leverage the diversity among multi-modal information to further refine the extracted features, reducing the redundancy in visual features, and we propose Grained Multi-modal Feature for Video Anomaly Detection (GMFVAD). Specifically, we generate more grained multi-modal feature based on the video snippet, which summarizes the main content, and text features based on the captions of original video will be introduced to further enhance the visual features of highlighted portions. Experiments show that the proposed GMFVAD achieves state-of-the-art performance on four mainly datasets. Ablation experiments also validate that the improvement of GMFVAD is due to the reduction of redundant information.
Authors: Jiayi Zou, Gengyun Jia, Bing-Kun Bao
Abstract: Visual Commonsense Reasoning (VCR) refers to answering questions and providing explanations based on images. While existing methods achieve high prediction accuracy, they often overlook bias in datasets and lack debiasing strategies. In this paper, our analysis reveals co-occurrence and statistical biases in both textual and visual data. We introduce the VCR-OOD datasets, comprising VCR-OOD-QA and VCR-OOD-VA subsets, which are designed to evaluate the generalization capabilities of models across two modalities. Furthermore, we analyze the causal graphs and prediction shortcuts in VCR and adopt a backdoor adjustment method to remove bias. Specifically, we create a dictionary based on the set of correct answers to eliminate prediction shortcuts. Experiments demonstrate the effectiveness of our debiasing method across different datasets.
Authors: Haodong Yang, Zhongling Huang, Shaojie Guo, Zhe Zhang, Gong Cheng, Junwei Han
Abstract: Deep learning models for complex-valued Synthetic Aperture Radar (CV-SAR) image recognition are fundamentally constrained by a representation trilemma under data-limited and domain-shift scenarios: the concurrent, yet conflicting, optimization of generalization, interpretability, and efficiency. Our work is motivated by the premise that the rich electromagnetic scattering features inherent in CV-SAR data hold the key to resolving this trilemma, yet they are insufficiently harnessed by conventional data-driven models. To this end, we introduce the Knowledge-Informed Neural Network (KINN), a lightweight framework built upon a novel "compression-aggregation-compression" architecture. The first stage performs a physics-guided compression, wherein a novel dictionary processor adaptively embeds physical priors, enabling a compact unfolding network to efficiently extract sparse, physically-grounded signatures. A subsequent aggregation module enriches these representations, followed by a final semantic compression stage that utilizes a compact classification head with self-distillation to learn maximally task-relevant and discriminative embeddings. We instantiate KINN in both CNN (0.7M) and Vision Transformer (0.95M) variants. Extensive evaluations on five SAR benchmarks confirm that KINN establishes a state-of-the-art in parameter-efficient recognition, offering exceptional generalization in data-scarce and out-of-distribution scenarios and tangible interpretability, thereby providing an effective solution to the representation trilemma and offering a new path for trustworthy AI in SAR image analysis.
Authors: Jiayi Zou, Chaofan Chen, Bing-Kun Bao, Changsheng Xu
Abstract: Egocentric Video Question Answering (Egocentric VideoQA) plays an important role in egocentric video understanding, which refers to answering questions based on first-person videos. Although existing methods have made progress through the paradigm of pre-training and fine-tuning, they ignore the unique challenges posed by the first-person perspective, such as understanding multiple events and recognizing hand-object interactions. To deal with these challenges, we propose a Dual-Modal Counterfactual Contrastive Construction (DMC$^3$) framework, which contains an egocentric videoqa baseline, a counterfactual sample construction module and a counterfactual sample-involved contrastive optimization. Specifically, We first develop a counterfactual sample construction module to generate positive and negative samples for textual and visual modalities through event description paraphrasing and core interaction mining, respectively. Then, We feed these samples together with the original samples into the baseline. Finally, in the counterfactual sample-involved contrastive optimization module, we apply contrastive loss to minimize the distance between the original sample features and the positive sample features, while maximizing the distance from the negative samples. Experiments show that our method achieve 52.51\% and 46.04\% on the \textit{normal} and \textit{indirect} splits of EgoTaskQA, and 13.2\% on QAEGO4D, both reaching the state-of-the-art performance.
Authors: Liangyu Chen, Hanzhang Zhou, Chenglin Cai, Jianan Zhang, Panrong Tong, Quyu Kong, Xu Zhang, Chen Liu, Yuqi Liu, Wenxuan Wang, Yue Wang, Qin Jin, Steven Hoi
Abstract: GUI grounding, which maps natural-language instructions to actionable UI elements, is a core capability of GUI agents. Prior works largely treats instructions as a static proxy for user intent, overlooking the impact of instruction diversity and quality on grounding performance. Through a careful investigation of existing grounding datasets, we find a 23.3% flaw rate in their instructions and show that inference-time exploitation of instruction diversity yields up to a substantial 76% relative performance improvement. In this paper, we introduce the Instruction-as-Reasoning paradigm, treating instructions as dynamic analytical pathways that offer distinct perspectives and enabling the model to select the most effective pathway during reasoning. To achieve this, we propose a two-stage training framework: supervised fine-tuning (SFT) on synthesized, diverse instructions to instill multi-perspective reasoning, followed by reinforcement learning (RL) to optimize pathway selection and composition. Our resulting models, UI-Ins-7B and UI-Ins-32B, achieve state-of-the-art results on five challenging grounding benchmarks and exhibit emergent reasoning, selectively composing and synthesizing novel instruction pathways at inference. In particular, UI-Ins-32B attains the best grounding accuracy, scoring 87.3% on UI-I2E-Bench, 57.0% on ScreenSpot-Pro, and 84.9% on MMBench-GUI L2. Furthermore, our model demonstrates strong agentic potential, achieving a 74.1% success rate on AndroidWorld using UI-Ins-7B as the executor. Our in-depth analysis reveals additional insights such as how reasoning can be formulated to enhance rather than hinder grounding performance, and how our method mitigates policy collapse in the SFT+RL framework. All code and model checkpoints will be publicly released in https://github.com/alibaba/UI-Ins.
Authors: Sauptik Dhar, Naveen Ramakrishnan, Michelle Munson
Abstract: Large Vision Language models have seen huge application in several sports use-cases recently. Most of these works have been targeted towards a limited subset of popular sports like soccer, cricket, basketball etc; focusing on generative tasks like visual question answering, highlight generation. This work analyzes the applicability of the modern video foundation models (both encoder and decoder) for a very niche but hugely popular dance sports - breakdance. Our results show that Video Encoder models continue to outperform state-of-the-art Video Language Models for prediction tasks. We provide insights on how to choose the encoder model and provide a thorough analysis into the workings of a finetuned decoder model for breakdance video classification.
Authors: LinFeng Li, Jian Zhao, Zepeng Yang, Yuhang Song, Bojun Lin, Tianle Zhang, Yuchen Yuan, Chi Zhang, Xuelong Li
Abstract: We present a winning solution to RoboSense 2025 Track 4: Cross-Modal Drone Navigation. The task retrieves the most relevant geo-referenced image from a large multi-platform corpus (satellite/drone/ground) given a natural-language query. Two obstacles are severe inter-platform heterogeneity and a domain gap between generic training descriptions and platform-specific test queries. We mitigate these with a domain-aligned preprocessing pipeline and a Mixture-of-Experts (MoE) framework: (i) platform-wise partitioning, satellite augmentation, and removal of orientation words; (ii) an LLM-based caption refinement pipeline to align textual semantics with the distinct visual characteristics of each platform. Using BGE-M3 (text) and EVA-CLIP (image), we train three platform experts using a progressive two-stage, hard-negative mining strategy to enhance discriminative power, and fuse their scores at inference. The system tops the official leaderboard, demonstrating robust cross-modal geo-localization under heterogeneous viewpoints.
Authors: Zelin Peng, Zhengqin Xu, Qingyang Liu, Xiaokang Yang, Wei Shen
Abstract: Multi-modal large language models (MLLMs) have emerged as a transformative approach for aligning visual and textual understanding. They typically require extremely high computational resources (e.g., thousands of GPUs) for training to achieve cross-modal alignment at multi-granularity levels. We argue that a key source of this inefficiency lies in the vision encoders they widely equip with, e.g., CLIP and SAM, which lack the alignment with language at multi-granularity levels. To address this issue, in this paper, we leverage hyperbolic space, which inherently models hierarchical levels and thus provides a principled framework for bridging the granularity gap between visual and textual modalities at an arbitrary granularity level. Concretely, we propose an efficient training paradigm for MLLMs, dubbed as HyperET, which can optimize visual representations to align with their textual counterparts at an arbitrary granularity level through dynamic hyperbolic radius adjustment in hyperbolic space. HyperET employs learnable matrices with M\"{o}bius multiplication operations, implemented via three effective configurations: diagonal scaling matrices, block-diagonal matrices, and banded matrices, providing a flexible yet efficient parametrization strategy. Comprehensive experiments across multiple MLLM benchmarks demonstrate that HyperET consistently improves both existing pre-training and fine-tuning MLLMs clearly with less than 1\% additional parameters.
Authors: Kangli Wang, Qianxi Yi, Yuqi Ye, Shihao Li, Wei Gao
Abstract: Generalization remains a critical challenge for deep learning-based point cloud geometry compression. We argue this stems from two key limitations: the lack of robust context models and the inefficient handling of out-of-distribution (OOD) data. To address both, we introduce AnyPcc, a universal point cloud compression framework. AnyPcc first employs a Universal Context Model that leverages priors from both spatial and channel-wise grouping to capture robust contextual dependencies. Second, our novel Instance-Adaptive Fine-Tuning (IAFT) strategy tackles OOD data by synergizing explicit and implicit compression paradigms. It fine-tunes a small subset of network weights for each instance and incorporates them into the bitstream, where the marginal bit cost of the weights is dwarfed by the resulting savings in geometry compression. Extensive experiments on a benchmark of 15 diverse datasets confirm that AnyPcc sets a new state-of-the-art in point cloud compression. Our code and datasets will be released to encourage reproducible research.
Authors: Seunghoon Lee, Jeongwoo Choi, Byunggwan Son, Jaehyeon Moon, Jeimin Jeon, Bumsub Ham
Abstract: We present in this paper a novel post-training quantization (PTQ) method, dubbed AccuQuant, for diffusion models. We show analytically and empirically that quantization errors for diffusion models are accumulated over denoising steps in a sampling process. To alleviate the error accumulation problem, AccuQuant minimizes the discrepancies between outputs of a full-precision diffusion model and its quantized version within a couple of denoising steps. That is, it simulates multiple denoising steps of a diffusion sampling process explicitly for quantization, accounting the accumulated errors over multiple denoising steps, which is in contrast to previous approaches to imitating a training process of diffusion models, namely, minimizing the discrepancies independently for each step. We also present an efficient implementation technique for AccuQuant, together with a novel objective, which reduces a memory complexity significantly from $\mathcal{O}(n)$ to $\mathcal{O}(1)$, where $n$ is the number of denoising steps. We demonstrate the efficacy and efficiency of AccuQuant across various tasks and diffusion models on standard benchmarks.
Authors: Yunpeng Bai, Haoxiang Li, Qixing Huang
Abstract: Diffusion Transformers (DiTs) have emerged as the dominant architecture for visual generation, powering state-of-the-art image and video models. By representing images as patch tokens with positional encodings (PEs), DiTs combine Transformer scalability with spatial and temporal inductive biases. In this work, we revisit how DiTs organize visual content and discover that patch tokens exhibit a surprising degree of independence: even when PEs are perturbed, DiTs still produce globally coherent outputs, indicating that spatial coherence is primarily governed by PEs. Motivated by this finding, we introduce the Positional Encoding Field (PE-Field), which extends positional encodings from the 2D plane to a structured 3D field. PE-Field incorporates depth-aware encodings for volumetric reasoning and hierarchical encodings for fine-grained sub-patch control, enabling DiTs to model geometry directly in 3D space. Our PE-Field-augmented DiT achieves state-of-the-art performance on single-image novel view synthesis and generalizes to controllable spatial image editing.
Authors: Qing Wang, Chong-Wah Ngo, Yu Cao, Ee-Peng Lim
Abstract: Existing approaches for image-to-recipe retrieval have the implicit assumption that a food image can fully capture the details textually documented in its recipe. However, a food image only reflects the visual outcome of a cooked dish and not the underlying cooking process. Consequently, learning cross-modal representations to bridge the modality gap between images and recipes tends to ignore subtle, recipe-specific details that are not visually apparent but are crucial for recipe retrieval. Specifically, the representations are biased to capture the dominant visual elements, resulting in difficulty in ranking similar recipes with subtle differences in use of ingredients and cooking methods. The bias in representation learning is expected to be more severe when the training data is mixed of images and recipes sourced from different cuisines. This paper proposes a novel causal approach that predicts the culinary elements potentially overlooked in images, while explicitly injecting these elements into cross-modal representation learning to mitigate biases. Experiments are conducted on the standard monolingual Recipe1M dataset and a newly curated multilingual multicultural cuisine dataset. The results indicate that the proposed causal representation learning is capable of uncovering subtle ingredients and cooking actions and achieves impressive retrieval performance on both monolingual and multilingual multicultural datasets.
Authors: Saif Ur Rehman Khan, Muhammad Nabeel Asim, Sebastian Vollmer, Andreas Dengel
Abstract: This paper presents the FuzzyDistillViT-MobileNet model, a novel approach for lung cancer (LC) classification, leveraging dynamic fuzzy logic-driven knowledge distillation (KD) to address uncertainty and complexity in disease diagnosis. Unlike traditional models that rely on static KD with fixed weights, our method dynamically adjusts the distillation weight using fuzzy logic, enabling the student model to focus on high-confidence regions while reducing attention to ambiguous areas. This dynamic adjustment improves the model ability to handle varying uncertainty levels across different regions of LC images. We employ the Vision Transformer (ViT-B32) as the instructor model, which effectively transfers knowledge to the student model, MobileNet, enhancing the student generalization capabilities. The training process is further optimized using a dynamic wait adjustment mechanism that adapts the training procedure for improved convergence and performance. To enhance image quality, we introduce pixel-level image fusion improvement techniques such as Gamma correction and Histogram Equalization. The processed images (Pix1 and Pix2) are fused using a wavelet-based fusion method to improve image resolution and feature preservation. This fusion method uses the wavedec2 function to standardize images to a 224x224 resolution, decompose them into multi-scale frequency components, and recursively average coefficients at each level for better feature representation. To address computational efficiency, Genetic Algorithm (GA) is used to select the most suitable pre-trained student model from a pool of 12 candidates, balancing model performance with computational cost. The model is evaluated on two datasets, including LC25000 histopathological images (99.16% accuracy) and IQOTH/NCCD CT-scan images (99.54% accuracy), demonstrating robustness across different imaging domains.
Authors: Kun Ouyang, Yuanxin Liu, Linli Yao, Yishuo Cai, Hao Zhou, Jie Zhou, Fandong Meng, Xu Sun
Abstract: Video reasoning, which requires multi-step deduction across frames, remains a major challenge for multimodal large language models (MLLMs). While reinforcement learning (RL)-based methods enhance reasoning capabilities, they often rely on text-only chains that yield ungrounded or hallucinated conclusions. Conversely, frame-retrieval approaches introduce visual grounding but still struggle with inaccurate evidence localization. To address these challenges, we present Conan, a framework for evidence-grounded multi-step video reasoning. Conan identifies contextual and evidence frames, reasons over cross-frame clues, and adaptively decides when to conclude or explore further. To achieve this, we (1) construct Conan-91K, a large-scale dataset of automatically generated reasoning traces that includes frame identification, evidence reasoning, and action decision, and (2) design a multi-stage progressive cold-start strategy combined with an Identification-Reasoning-Action (AIR) RLVR training framework to jointly enhance multi-step visual reasoning. Extensive experiments on six multi-step reasoning benchmarks demonstrate that Conan surpasses the baseline Qwen2.5-VL-7B-Instruct by an average of over 10% in accuracy, achieving state-of-the-art performance. Furthermore, Conan generalizes effectively to long-video understanding tasks, validating its strong scalability and robustness.
Authors: Alexandre Fournier-Montgieux, Herv\'e Le Borgne, Adrian Popescu, Bertrand Luvison
Abstract: Fairness evaluation in face analysis systems (FAS) typically depends on automatic demographic attribute inference (DAI), which itself relies on predefined demographic segmentation. However, the validity of fairness auditing hinges on the reliability of the DAI process. We begin by providing a theoretical motivation for this dependency, showing that improved DAI reliability leads to less biased and lower-variance estimates of FAS fairness. To address this, we propose a fully reproducible DAI pipeline that replaces conventional end-to-end training with a modular transfer learning approach. Our design integrates pretrained face recognition encoders with non-linear classification heads. We audit this pipeline across three dimensions: accuracy, fairness, and a newly introduced notion of robustness, defined via intra-identity consistency. The proposed robustness metric is applicable to any demographic segmentation scheme. We benchmark the pipeline on gender and ethnicity inference across multiple datasets and training setups. Our results show that the proposed method outperforms strong baselines, particularly on ethnicity, which is the more challenging attribute. To promote transparency and reproducibility, we will publicly release the training dataset metadata, full codebase, pretrained models, and evaluation toolkit. This work contributes a reliable foundation for demographic inference in fairness auditing.
Authors: Yixiong Yang, Tao Wu, Senmao Li, Shiqi Yang, Yaxing Wang, Joost van de Weijer, Kai Wang
Abstract: Recent advances in accelerating text-to-image (T2I) diffusion models have enabled the synthesis of high-fidelity images even in a single step. However, personalizing these models to incorporate novel concepts remains a challenge due to the limited capacity of one-step models to capture new concept distributions effectively. We propose a bidirectional concept distillation framework, EchoDistill, to enable one-step diffusion personalization (1-SDP). Our approach involves an end-to-end training process where a multi-step diffusion model (teacher) and a one-step diffusion model (student) are trained simultaneously. The concept is first distilled from the teacher model to the student, and then echoed back from the student to the teacher. During the EchoDistill, we share the text encoder between the two models to ensure consistent semantic understanding. Following this, the student model is optimized with adversarial losses to align with the real image distribution and with alignment losses to maintain consistency with the teacher's output. Furthermore, we introduce the bidirectional echoing refinement strategy, wherein the student model leverages its faster generation capability to feedback to the teacher model. This bidirectional concept distillation mechanism not only enhances the student ability to personalize novel concepts but also improves the generative quality of the teacher model. Our experiments demonstrate that this collaborative framework significantly outperforms existing personalization methods over the 1-SDP setup, establishing a novel paradigm for rapid and effective personalization in T2I diffusion models.
Authors: Xiaohan Lan, Fanfan Liu, Haibo Qiu, Siqi Yang, Delian Ruan, Peng Shi, Lin Ma
Abstract: Inspired by recent advancements in LLM reasoning, the field of multimodal reasoning has seen remarkable progress, achieving significant performance gains on intricate tasks such as mathematical problem-solving. Despite this progress, current multimodal large reasoning models exhibit two key limitations. They tend to employ computationally expensive reasoning even for simple queries, leading to inefficiency. Furthermore, this focus on specialized reasoning often impairs their broader, more general understanding capabilities. In this paper, we propose Metis-HOME: a Hybrid Optimized Mixture-of-Experts framework designed to address this trade-off. Metis-HOME enables a ''Hybrid Thinking'' paradigm by structuring the original dense model into two distinct expert branches: a thinking branch tailored for complex, multi-step reasoning, and a non-thinking branch optimized for rapid, direct inference on tasks like general VQA and OCR. A lightweight, trainable router dynamically allocates queries to the most suitable expert. We instantiate Metis-HOME by adapting the Qwen2.5-VL-7B into an MoE architecture. Comprehensive evaluations reveal that our approach not only substantially enhances complex reasoning abilities but also improves the model's general capabilities, reversing the degradation trend observed in other reasoning-specialized models. Our work establishes a new paradigm for building powerful and versatile MLLMs, effectively resolving the prevalent reasoning-vs-generalization dilemma.
Authors: Lixiong Qin, Yang Zhang, Mei Wang, Jiani Hu, Weihong Deng, Weiran Xu
Abstract: The advancement of Multimodal Large Language Models (MLLMs) has bridged the gap between vision and language tasks, enabling the implementation of Explainable DeepFake Analysis (XDFA). However, current methods suffer from a lack of fine-grained awareness: the description of artifacts in data annotation is unreliable and coarse-grained, and the models fail to support the output of connections between textual forgery explanations and the visual evidence of artifacts, as well as the input of queries for arbitrary facial regions. As a result, their responses are not sufficiently grounded in Face Visual Context (Facext). To address this limitation, we propose the Fake-in-Facext (FiFa) framework, with contributions focusing on data annotation and model construction. We first define a Facial Image Concept Tree (FICT) to divide facial images into fine-grained regional concepts, thereby obtaining a more reliable data annotation pipeline, FiFa-Annotator, for forgery explanation. Based on this dedicated data annotation, we introduce a novel Artifact-Grounding Explanation (AGE) task, which generates textual forgery explanations interleaved with segmentation masks of manipulated artifacts. We propose a unified multi-task learning architecture, FiFa-MLLM, to simultaneously support abundant multimodal inputs and outputs for fine-grained Explainable DeepFake Analysis. With multiple auxiliary supervision tasks, FiFa-MLLM can outperform strong baselines on the AGE task and achieve SOTA performance on existing XDFA datasets. The code and data will be made open-source at https://github.com/lxq1000/Fake-in-Facext.
Authors: Guillermo Carbajal, Andr\'es Almansa, Pablo Mus\'e
Abstract: Motion blur caused by camera shake, particularly under large or rotational movements, remains a major challenge in image restoration. We propose a deep learning framework that jointly estimates the latent sharp image and the underlying camera motion trajectory from a single blurry image. Our method leverages the Projective Motion Blur Model (PMBM), implemented efficiently using a differentiable blur creation module compatible with modern networks. A neural network predicts a full 3D rotation trajectory, which guides a model-based restoration network trained end-to-end. This modular architecture provides interpretability by revealing the camera motion that produced the blur. Moreover, this trajectory enables the reconstruction of the sequence of sharp images that generated the observed blurry image. To further refine results, we optimize the trajectory post-inference via a reblur loss, improving consistency between the blurry input and the restored output. Extensive experiments show that our method achieves state-of-the-art performance on both synthetic and real datasets, particularly in cases with severe or spatially variant blur, where end-to-end deblurring networks struggle. Code and trained models are available at https://github.com/GuillermoCarbajal/Blur2Seq/
Authors: Marziyeh Bamdad, Hans-Peter Hutter, Alireza Darvishy
Abstract: Despite advancements in SLAM technologies, robust operation under challenging conditions such as low-texture, motion-blur, or challenging lighting remains an open challenge. Such conditions are common in applications such as assistive navigation for the visually impaired. These challenges undermine localization accuracy and tracking stability, reducing navigation reliability and safety. To overcome these limitations, we present SELM-SLAM3, a deep learning-enhanced visual SLAM framework that integrates SuperPoint and LightGlue for robust feature extraction and matching. We evaluated our framework using TUM RGB-D, ICL-NUIM, and TartanAir datasets, which feature diverse and challenging scenarios. SELM-SLAM3 outperforms conventional ORB-SLAM3 by an average of 87.84% and exceeds state-of-the-art RGB-D SLAM systems by 36.77%. Our framework demonstrates enhanced performance under challenging conditions, such as low-texture scenes and fast motion, providing a reliable platform for developing navigation aids for the visually impaired.
Authors: Fuchen Li, Yansong Du, Wenbo Cheng, Xiaoxia Zhou, Sen Yin
Abstract: Consumer-grade camera systems often struggle to maintain stable image quality under complex illumination conditions such as low light, high dynamic range, and backlighting, as well as spatial color temperature variation. These issues lead to underexposure, color casts, and tonal inconsistency, which degrade the performance of downstream vision tasks. To address this, we propose ACamera-Net, a lightweight and scene-adaptive camera parameter adjustment network that directly predicts optimal exposure and white balance from RAW inputs. The framework consists of two modules: ACamera-Exposure, which estimates ISO to alleviate underexposure and contrast loss, and ACamera-Color, which predicts correlated color temperature and gain factors for improved color consistency. Optimized for real-time inference on edge devices, ACamera-Net can be seamlessly integrated into imaging pipelines. Trained on diverse real-world data with annotated references, the model generalizes well across lighting conditions. Extensive experiments demonstrate that ACamera-Net consistently enhances image quality and stabilizes perception outputs, outperforming conventional auto modes and lightweight baselines without relying on additional image enhancement modules.
Authors: Xiaohan Sun, Carol O'Sullivan
Abstract: In this paper, we investigate how users perceive the visual quality of crowd character representations at different levels of detail (LoD) and viewing distances. Each representation: geometric meshes, image-based impostors, Neural Radiance Fields (NeRFs), and 3D Gaussians, exhibits distinct trade-offs between visual fidelity and computational performance. Our qualitative and quantitative results provide insights to guide the design of perceptually optimized LoD strategies for crowd rendering.
Authors: Ding Zou, Feifan Wang, Mengyu Ge, Siyuan Fan, Zongbing Zhang, Wei Chen, Lingfeng Wang, Zhongyou Hu, Wenrui Yan, Zhengwei Gao, Hao Wang, Weizhao Jin, Yu Zhang, Hainan Zhao, Mingliang Zhang, Xianxian Xi, Yaru Zhang, Wenyuan Li, Zhengguang Gao, Yurui Zhu
Abstract: The realization of Artificial General Intelligence (AGI) necessitates Embodied AI agents capable of robust spatial perception, effective task planning, and adaptive execution in physical environments. However, current large language models (LLMs) and multimodal LLMs (MLLMs) for embodied tasks suffer from key limitations, including a significant gap between model design and agent requirements, an unavoidable trade-off between real-time latency and performance, and the use of unauthentic, offline evaluation metrics. To address these challenges, we propose EmbodiedBrain, a novel vision-language foundation model available in both 7B and 32B parameter sizes. Our framework features an agent-aligned data structure and employs a powerful training methodology that integrates large-scale Supervised Fine-Tuning (SFT) with Step-Augumented Group Relative Policy Optimization (Step-GRPO), which boosts long-horizon task success by integrating preceding steps as Guided Precursors. Furthermore, we incorporate a comprehensive reward system, including a Generative Reward Model (GRM) accelerated at the infrastructure level, to improve training efficiency. For enable thorough validation, we establish a three-part evaluation system encompassing General, Planning, and End-to-End Simulation Benchmarks, highlighted by the proposal and open-sourcing of a novel, challenging simulation environment. Experimental results demonstrate that EmbodiedBrain achieves superior performance across all metrics, establishing a new state-of-the-art for embodied foundation models. Towards paving the way for the next generation of generalist embodied agents, we open-source all of our data, model weight, and evaluating methods, which are available at https://zterobot.github.io/EmbodiedBrain.github.io.
Authors: Jiahao Meng, Xiangtai Li, Haochen Wang, Yue Tan, Tao Zhang, Lingdong Kong, Yunhai Tong, Anran Wang, Zhiyang Teng, Yujing Wang, Zhuochen Wang
Abstract: Most video reasoning models only generate textual reasoning traces without indicating when and where key evidence appears. Recent models such as OpenAI-o3 have sparked wide interest in evidence-centered reasoning for images, yet extending this ability to videos is more challenging, as it requires joint temporal tracking and spatial localization across dynamic scenes. We introduce Open-o3 Video, a non-agent framework that integrates explicit spatio-temporal evidence into video reasoning, and carefully collect training data and design training strategies to address the aforementioned challenges. The model highlights key timestamps, objects, and bounding boxes alongside its answers, allowing reasoning to be grounded in concrete visual observations. To enable this functionality, we first curate and build two high-quality datasets, STGR-CoT-30k for SFT and STGR-RL-36k for RL, with carefully constructed temporal and spatial annotations, since most existing datasets offer either temporal spans for videos or spatial boxes on images, lacking unified spatio-temporal supervision and reasoning traces. Then, we adopt a cold-start reinforcement learning strategy with multiple specially designed rewards that jointly encourage answer accuracy, temporal alignment, and spatial precision. On V-STAR benchmark, Open-o3 Video achieves state-of-the-art performance, raising mAM by 14.4% and mLGM by 24.2% on the Qwen2.5-VL baseline. Consistent improvements are also observed on a broad range of video understanding benchmarks, including VideoMME, WorldSense, VideoMMMU, and TVGBench. Beyond accuracy, the reasoning traces produced by Open-o3 Video also provide valuable signals for test-time scaling, enabling confidence-aware verification and improving answer reliability.
Authors: Muhammad Atif Butt, Alexandra Gomez-Villa, Tao Wu, Javier Vazquez-Corral, Joost Van De Weijer, Kai Wang
Abstract: Recent years have seen impressive advances in text-to-image generation, with image generative or unified models producing high-quality images from text. Yet these models still struggle with fine-grained color controllability, often failing to accurately match colors specified in text prompts. While existing benchmarks evaluate compositional reasoning and prompt adherence, none systematically assess color precision. Color is fundamental to human visual perception and communication, critical for applications from art to design workflows requiring brand consistency. However, current benchmarks either neglect color or rely on coarse assessments, missing key capabilities such as interpreting RGB values or aligning with human expectations. To this end, we propose GenColorBench, the first comprehensive benchmark for text-to-image color generation, grounded in color systems like ISCC-NBS and CSS3/X11, including numerical colors which are absent elsewhere. With 44K color-focused prompts covering 400+ colors, it reveals models' true capabilities via perceptual and automated assessments. Evaluations of popular text-to-image models using GenColorBench show performance variations, highlighting which color conventions models understand best and identifying failure modes. Our GenColorBench assessments will guide improvements in precise color generation. The benchmark will be made public upon acceptance.
Authors: Ziyu Ye, Chen Ju, Chaofan Ma, Xiaoyun Zhang
Abstract: Deep learning models have achieved great success on various vision challenges, but a well-trained model would face drastic performance degradation when applied to unseen data. Since the model is sensitive to domain shift, unsupervised domain adaptation attempts to reduce the domain gap and avoid costly annotation of unseen domains. This paper proposes a novel framework for cross-modality segmentation via similarity-based prototypes. In specific, we learn class-wise prototypes within an embedding space, then introduce a similarity constraint to make these prototypes representative for each semantic class while separable from different classes. Moreover, we use dictionaries to store prototypes extracted from different images, which prevents the class-missing problem and enables the contrastive learning of prototypes, and further improves performance. Extensive experiments show that our method achieves better results than other state-of-the-art methods.
Authors: Mark He Huang, Lin Geng Foo, Christian Theobalt, Ying Sun, De Wen Soh
Abstract: Free-moving object reconstruction from monocular video remains challenging, particularly without reliable pose or depth cues and under arbitrary object motion. We introduce OnlineSplatter, a novel online feed-forward framework generating high-quality, object-centric 3D Gaussians directly from RGB frames without requiring camera pose, depth priors, or bundle optimization. Our approach anchors reconstruction using the first frame and progressively refines the object representation through a dense Gaussian primitive field, maintaining constant computational cost regardless of video sequence length. Our core contribution is a dual-key memory module combining latent appearance-geometry keys with explicit directional keys, robustly fusing current frame features with temporally aggregated object states. This design enables effective handling of free-moving objects via spatial-guided memory readout and an efficient sparsification mechanism, ensuring comprehensive yet compact object coverage. Evaluations on real-world datasets demonstrate that OnlineSplatter significantly outperforms state-of-the-art pose-free reconstruction baselines, consistently improving with more observations while maintaining constant memory and runtime.
Authors: Yuan Sheng, Yanbin Hao, Chenxu Li, Shuo Wang, Xiangnan He
Abstract: Long video understanding remains challenging due to its complex, diverse, and temporally scattered content. Although video large language models (Video-LLMs) can process videos lasting tens of minutes, applying them to truly long sequences is computationally prohibitive and often leads to unfocused or inconsistent reasoning. A promising solution is to select only the most informative frames, yet existing approaches typically ignore temporal dependencies or rely on unimodal evidence, limiting their ability to provide complete and query-relevant context. We propose a Semantic-Visual Consensus Evidence Selection (SeViCES) framework for effective and reliable long video understanding. SeViCES is training-free and model-agnostic, and introduces two key components. The Semantic-Visual Consensus Frame Selection (SVCFS) module selects frames through (1) a temporal-aware semantic branch that leverages LLM reasoning over captions, and (2) a cluster-guided visual branch that aligns embeddings with semantic scores via mutual information. The Answer Consensus Refinement (ACR) module further resolves inconsistencies between semantic- and visual-based predictions by fusing evidence and constraining the answer space. Extensive experiments on long video understanding benchmarks show that SeViCES consistently outperforms state-of-the-art methods in both accuracy and robustness, demonstrating the importance of consensus-driven evidence selection for Video-LLMs.
Authors: Zhenhuan Zhou, Jingbo Zhu, Yuchen Zhang, Xiaohang Guan, Peng Wang, Tao Li
Abstract: Efficient analysis and processing of dental images are crucial for dentists to achieve accurate diagnosis and optimal treatment planning. However, dental imaging inherently poses several challenges, such as low contrast, metallic artifacts, and variations in projection angles. Combined with the subjectivity arising from differences in clinicians' expertise, manual interpretation often proves time-consuming and prone to inconsistency. Artificial intelligence (AI)-based automated dental image analysis (DIA) offers a promising solution to these issues and has become an integral part of computer-aided dental diagnosis and treatment. Among various AI technologies, deep learning (DL) stands out as the most widely applied and influential approach due to its superior feature extraction and representation capabilities. To comprehensively summarize recent progress in this field, we focus on the two fundamental aspects of DL research-datasets and models. In this paper, we systematically review 260 studies on DL applications in DIA, including 49 papers on publicly available dental datasets and 211 papers on DL-based algorithms. We first introduce the basic concepts of dental imaging and summarize the characteristics and acquisition methods of existing datasets. Then, we present the foundational techniques of DL and categorize relevant models and algorithms according to different DIA tasks, analyzing their network architectures, optimization strategies, training methods, and performance. Furthermore, we summarize commonly used training and evaluation metrics in the DIA domain. Finally, we discuss the current challenges of existing research and outline potential future directions. We hope that this work provides a valuable and systematic reference for researchers in this field. All supplementary materials and detailed comparison tables will be made publicly available on GitHub.
Authors: Ibrahim Ethem Hamamci, Sezgin Er, Suprosanna Shit, Hadrien Reynaud, Dong Yang, Pengfei Guo, Marc Edgar, Daguang Xu, Bernhard Kainz, Bjoern Menze
Abstract: Recent progress in vision-language modeling for 3D medical imaging has been fueled by large-scale computed tomography (CT) corpora with paired free-text reports, stronger architectures, and powerful pretrained models. This has enabled applications such as automated report generation and text-conditioned 3D image synthesis. Yet, current approaches struggle with high-resolution, long-sequence volumes: contrastive pretraining often yields vision encoders that are misaligned with clinical language, and slice-wise tokenization blurs fine anatomy, reducing diagnostic performance on downstream tasks. We introduce BTB3D (Better Tokens for Better 3D), a causal convolutional encoder-decoder that unifies 2D and 3D training and inference while producing compact, frequency-aware volumetric tokens. A three-stage training curriculum enables (i) local reconstruction, (ii) overlapping-window tiling, and (iii) long-context decoder refinement, during which the model learns from short slice excerpts yet generalizes to scans exceeding 300 slices without additional memory overhead. BTB3D sets a new state-of-the-art on two key tasks: it improves BLEU scores and increases clinical F1 by 40% over CT2Rep, CT-CHAT, and Merlin for report generation; and it reduces FID by 75% and halves FVD compared to GenerateCT and MedSyn for text-to-CT synthesis, producing anatomically consistent 512*512*241 volumes. These results confirm that precise three-dimensional tokenization, rather than larger language backbones alone, is essential for scalable vision-language modeling in 3D medical imaging. The codebase is available at: https://github.com/ibrahimethemhamamci/BTB3D
Authors: Chen Zhao, En Ci, Yunzhe Xu, Tiehan Fan, Shanyan Guan, Yanhao Ge, Jian Yang, Ying Tai
Abstract: Ultra-high-resolution (UHR) text-to-image (T2I) generation has seen notable progress. However, two key challenges remain : 1) the absence of a large-scale high-quality UHR T2I dataset, and (2) the neglect of tailored training strategies for fine-grained detail synthesis in UHR scenarios. To tackle the first challenge, we introduce \textbf{UltraHR-100K}, a high-quality dataset of 100K UHR images with rich captions, offering diverse content and strong visual fidelity. Each image exceeds 3K resolution and is rigorously curated based on detail richness, content complexity, and aesthetic quality. To tackle the second challenge, we propose a frequency-aware post-training method that enhances fine-detail generation in T2I diffusion models. Specifically, we design (i) \textit{Detail-Oriented Timestep Sampling (DOTS)} to focus learning on detail-critical denoising steps, and (ii) \textit{Soft-Weighting Frequency Regularization (SWFR)}, which leverages Discrete Fourier Transform (DFT) to softly constrain frequency components, encouraging high-frequency detail preservation. Extensive experiments on our proposed UltraHR-eval4K benchmarks demonstrate that our approach significantly improves the fine-grained detail quality and overall fidelity of UHR image generation. The code is available at \href{https://github.com/NJU-PCALab/UltraHR-100k}{here}.
Authors: Debojyoti Ghosh, Adrijit Goswami
Abstract: Accurate waste classification is vital for achieving sustainable waste management and reducing the environmental footprint of urbanization. Misclassification of recyclable materials contributes to landfill accumulation, inefficient recycling, and increased greenhouse gas emissions. To address these issues, this study introduces HybridSOMSpikeNet, a hybrid deep learning framework that integrates convolutional feature extraction, differentiable self-organization, and spiking-inspired temporal processing to enable intelligent and energy-efficient waste classification. The proposed model employs a pre-trained ResNet-152 backbone to extract deep spatial representations, followed by a Differentiable Soft Self-Organizing Map (Soft-SOM) that enhances topological clustering and interpretability. A spiking neural head accumulates temporal activations over discrete time steps, improving robustness and generalization. Trained on a ten-class waste dataset, HybridSOMSpikeNet achieved a test accuracy of 97.39%, outperforming several state-of-the-art architectures while maintaining a lightweight computational profile suitable for real-world deployment. Beyond its technical innovations, the framework provides tangible environmental benefits. By enabling precise and automated waste segregation, it supports higher recycling efficiency, reduces contamination in recyclable streams, and minimizes the ecological and operational costs of waste processing. The approach aligns with global sustainability priorities, particularly the United Nations Sustainable Development Goals (SDG 11 and SDG 12), by contributing to cleaner cities, circular economy initiatives, and intelligent environmental management systems.
Authors: Jinhee Kim, Jae Jun An, Kang Eun Jeon, Jong Hwan Ko
Abstract: Multi-bit quantization networks enable flexible deployment of deep neural networks by supporting multiple precision levels within a single model. However, existing approaches suffer from significant training overhead as full-dataset updates are repeated for each supported bit-width, resulting in a cost that scales linearly with the number of precisions. Additionally, extra fine-tuning stages are often required to support additional or intermediate precision options, further compounding the overall training burden. To address this issue, we propose two techniques that greatly reduce the training overhead without compromising model utility: (i) Weight bias correction enables shared batch normalization and eliminates the need for fine-tuning by neutralizing quantization-induced bias across bit-widths and aligning activation distributions; and (ii) Bit-wise coreset sampling strategy allows each child model to train on a compact, informative subset selected via gradient-based importance scores by exploiting the implicit knowledge transfer phenomenon. Experiments on CIFAR-10/100, TinyImageNet, and ImageNet-1K with both ResNet and ViT architectures demonstrate that our method achieves competitive or superior accuracy while reducing training time up to 7.88x. Our code is released at https://github.com/a2jinhee/EMQNet_jk.
Authors: Jing Bi, Guangyu Sun, Ali Vosoughi, Chen Chen, Chenliang Xu
Abstract: Multimodal large language models (MLLMs) that integrate visual and textual reasoning leverage chain-of-thought (CoT) prompting to tackle complex visual tasks, yet continue to exhibit visual hallucinations and an over-reliance on textual priors. We present a systematic diagnosis of state-of-the-art vision-language models using a three-stage evaluation framework, uncovering key failure modes. To address these, we propose an agent-based architecture that combines LLM reasoning with lightweight visual modules, enabling fine-grained analysis and iterative refinement of reasoning chains. Our results highlight future visual reasoning models should focus on integrating a broader set of specialized tools for analyzing visual content. Our system achieves significant gains (+10.3 on MMMU, +6.0 on MathVista over a 7B baseline), matching or surpassing much larger models. We will release our framework and evaluation suite to facilitate future research.
Authors: Xuyang Liu, Xiyan Gui, Yuchao Zhang, Linfeng Zhang
Abstract: Recent large vision-language models (LVLMs) demonstrate remarkable capabilities in processing extended multi-modal sequences, yet the resulting key-value (KV) cache expansion creates a critical memory bottleneck that fundamentally limits deployment scalability. While existing KV cache compression methods focus on retaining high-importance KV pairs to minimize storage, they often overlook the modality-specific semantic redundancy patterns that emerge distinctively in multi-modal KV caches. In this work, we first analyze how, beyond simple importance, the KV cache in LVLMs exhibits varying levels of redundancy across attention heads. We show that relying solely on importance can only cover a subset of the full KV cache information distribution, leading to potential loss of semantic coverage. To address this, we propose \texttt{MixKV}, a novel method that mixes importance with diversity for optimized KV cache compression in LVLMs. \texttt{MixKV} adapts to head-wise semantic redundancy, selectively balancing diversity and importance when compressing KV pairs. Extensive experiments demonstrate that \texttt{MixKV} consistently enhances existing methods across multiple LVLMs. Under extreme compression (budget=64), \texttt{MixKV} improves baseline methods by an average of \textbf{5.1\%} across five multi-modal understanding benchmarks and achieves remarkable gains of \textbf{8.0\%} and \textbf{9.0\%} for SnapKV and AdaKV on GUI grounding tasks, all while maintaining comparable inference efficiency. Furthermore, \texttt{MixKV} extends seamlessly to LLMs with comparable performance gains. Our code is available at \href{https://github.com/xuyang-liu16/MixKV}{\textcolor{citeblue}{https://github.com/xuyang-liu16/MixKV}}.
URLs: https://github.com/xuyang-liu16/MixKV, https://github.com/xuyang-liu16/MixKV
Authors: Samuel Soutullo, Miguel Yermo, David L. Vilari\~no, \'Oscar G. Lorenzo, Jos\'e C. Cabaleiro, Francisco F. Rivera
Abstract: 3D LiDAR sensors are essential for autonomous navigation, environmental monitoring, and precision mapping in remote sensing applications. To efficiently process the massive point clouds generated by these sensors, LiDAR data is often projected into 2D range images that organize points by their angular positions and distances. While these range image representations enable efficient processing, conventional projection methods suffer from fundamental geometric inconsistencies that cause irreversible information loss, compromising high-fidelity applications. We present ALICE-LRI (Automatic LiDAR Intrinsic Calibration Estimation for Lossless Range Images), the first general, sensor-agnostic method that achieves lossless range image generation from spinning LiDAR point clouds without requiring manufacturer metadata or calibration files. Our algorithm automatically reverse-engineers the intrinsic geometry of any spinning LiDAR sensor by inferring critical parameters including laser beam configuration, angular distributions, and per-beam calibration corrections, enabling lossless projection and complete point cloud reconstruction with zero point loss. Comprehensive evaluation across the complete KITTI and DurLAR datasets demonstrates that ALICE-LRI achieves perfect point preservation, with zero points lost across all point clouds. Geometric accuracy is maintained well within sensor precision limits, establishing geometric losslessness with real-time performance. We also present a compression case study that validates substantial downstream benefits, demonstrating significant quality improvements in practical applications. This paradigm shift from approximate to lossless LiDAR projections opens new possibilities for high-precision remote sensing applications requiring complete geometric preservation.
Authors: Jiacheng Chen, Ziyu Jiang, Mingfu Liang, Bingbing Zhuang, Jong-Chyi Su, Sparsh Garg, Ying Wu, Manmohan Chandraker
Abstract: This paper proposes AutoScape, a long-horizon driving scene generation framework. At its core is a novel RGB-D diffusion model that iteratively generates sparse, geometrically consistent keyframes, serving as reliable anchors for the scene's appearance and geometry. To maintain long-range geometric consistency, the model 1) jointly handles image and depth in a shared latent space, 2) explicitly conditions on the existing scene geometry (i.e., rendered point clouds) from previously generated keyframes, and 3) steers the sampling process with a warp-consistent guidance. Given high-quality RGB-D keyframes, a video diffusion model then interpolates between them to produce dense and coherent video frames. AutoScape generates realistic and geometrically consistent driving videos of over 20 seconds, improving the long-horizon FID and FVD scores over the prior state-of-the-art by 48.6\% and 43.0\%, respectively.
Authors: Nima Torbati, Anastasia Meshcheryakova, Ramona Woitek, Diana Mechtcheriakova, Amirreza Mahbod
Abstract: Automated histopathological image analysis plays a vital role in computer-aided diagnosis of various diseases. Among developed algorithms, deep learning-based approaches have demonstrated excellent performance in multiple tasks, including semantic tissue segmentation in histological images. In this study, we propose a novel approach based on attention-driven feature fusion of convolutional neural networks (CNNs) and vision transformers (ViTs) within a unified dual-encoder model to improve semantic segmentation performance. Evaluation on two publicly available datasets showed that our model achieved {\mu}IoU/{\mu}Dice scores of 76.79%/86.87% on the GCPS dataset and 64.93%/76.60% on the PUMA dataset, outperforming state-of-the-art and baseline benchmarks. The implementation of our method is publicly available in a GitHub repository: https://github.com/NimaTorbati/ACS-SegNet
Authors: Noam Issachar, Guy Yariv, Sagie Benaim, Yossi Adi, Dani Lischinski, Raanan Fattal
Abstract: Diffusion Transformer models can generate images with remarkable fidelity and detail, yet training them at ultra-high resolutions remains extremely costly due to the self-attention mechanism's quadratic scaling with the number of image tokens. In this paper, we introduce Dynamic Position Extrapolation (DyPE), a novel, training-free method that enables pre-trained diffusion transformers to synthesize images at resolutions far beyond their training data, with no additional sampling cost. DyPE takes advantage of the spectral progression inherent to the diffusion process, where low-frequency structures converge early, while high-frequencies take more steps to resolve. Specifically, DyPE dynamically adjusts the model's positional encoding at each diffusion step, matching their frequency spectrum with the current stage of the generative process. This approach allows us to generate images at resolutions that exceed the training resolution dramatically, e.g., 16 million pixels using FLUX. On multiple benchmarks, DyPE consistently improves performance and achieves state-of-the-art fidelity in ultra-high-resolution image generation, with gains becoming even more pronounced at higher resolutions. Project page is available at https://noamissachar.github.io/DyPE/.
Authors: Huijie Zhang, Aliaksandr Siarohin, Willi Menapace, Michael Vasilkovsky, Sergey Tulyakov, Qing Qu, Ivan Skorokhodov
Abstract: MeanFlow has recently emerged as a powerful framework for few-step generative modeling trained from scratch, but its success is not yet fully understood. In this work, we show that the MeanFlow objective naturally decomposes into two parts: trajectory flow matching and trajectory consistency. Through gradient analysis, we find that these terms are strongly negatively correlated, causing optimization conflict and slow convergence. Motivated by these insights, we introduce $\alpha$-Flow, a broad family of objectives that unifies trajectory flow matching, Shortcut Model, and MeanFlow under one formulation. By adopting a curriculum strategy that smoothly anneals from trajectory flow matching to MeanFlow, $\alpha$-Flow disentangles the conflicting objectives, and achieves better convergence. When trained from scratch on class-conditional ImageNet-1K 256x256 with vanilla DiT backbones, $\alpha$-Flow consistently outperforms MeanFlow across scales and settings. Our largest $\alpha$-Flow-XL/2+ model achieves new state-of-the-art results using vanilla DiT backbones, with FID scores of 2.58 (1-NFE) and 2.15 (2-NFE).
Authors: Binbin Huang, Haobin Duan, Yiqun Zhao, Zibo Zhao, Yi Ma, Shenghua Gao
Abstract: This work proposes a new generation-based 3D reconstruction method, named Cupid, that accurately infers the camera pose, 3D shape, and texture of an object from a single 2D image. Cupid casts 3D reconstruction as a conditional sampling process from a learned distribution of 3D objects, and it jointly generates voxels and pixel-voxel correspondences, enabling robust pose and shape estimation under a unified generative framework. By representing both input camera poses and 3D shape as a distribution in a shared 3D latent space, Cupid adopts a two-stage flow matching pipeline: (1) a coarse stage that produces initial 3D geometry with associated 2D projections for pose recovery; and (2) a refinement stage that integrates pose-aligned image features to enhance structural fidelity and appearance details. Extensive experiments demonstrate Cupid outperforms leading 3D reconstruction methods with an over 3 dB PSNR gain and an over 10% Chamfer Distance reduction, while matching monocular estimators on pose accuracy and delivering superior visual fidelity over baseline 3D generative models. For an immersive view of the 3D results generated by Cupid, please visit cupid3d.github.io.
Authors: Lei Cheng, Siyang Cao
Abstract: This paper presents a Multi-Object Tracking (MOT) framework that fuses radar and camera data to enhance tracking efficiency while minimizing manual interventions. Contrary to many studies that underutilize radar and assign it a supplementary role--despite its capability to provide accurate range/depth information of targets in a world 3D coordinate system--our approach positions radar in a crucial role. Meanwhile, this paper utilizes common features to enable online calibration to autonomously associate detections from radar and camera. The main contributions of this work include: (1) the development of a radar-camera fusion MOT framework that exploits online radar-camera calibration to simplify the integration of detection results from these two sensors, (2) the utilization of common features between radar and camera data to accurately derive real-world positions of detected objects, and (3) the adoption of feature matching and category-consistency checking to surpass the limitations of mere position matching in enhancing sensor association accuracy. To the best of our knowledge, we are the first to investigate the integration of radar-camera common features and their use in online calibration for achieving MOT. The efficacy of our framework is demonstrated by its ability to streamline the radar-camera mapping process and improve tracking precision, as evidenced by real-world experiments conducted in both controlled environments and actual traffic scenarios. Code is available at https://github.com/radar-lab/Radar_Camera_MOT
Authors: Xiaolong Wang, Lixiang Ru, Ziyuan Huang, Kaixiang Ji, Dandan Zheng, Jingdong Chen, Jun Zhou
Abstract: We propose a novel AutoRegressive Generation-based paradigm for image Segmentation (ARGenSeg), achieving multimodal understanding and pixel-level perception within a unified framework. Prior works integrating image segmentation into multimodal large language models (MLLMs) typically employ either boundary points representation or dedicated segmentation heads. These methods rely on discrete representations or semantic prompts fed into task-specific decoders, which limits the ability of the MLLM to capture fine-grained visual details. To address these challenges, we introduce a segmentation framework for MLLM based on image generation, which naturally produces dense masks for target objects. We leverage MLLM to output visual tokens and detokenize them into images using an universal VQ-VAE, making the segmentation fully dependent on the pixel-level understanding of the MLLM. To reduce inference latency, we employ a next-scale-prediction strategy to generate required visual tokens in parallel. Extensive experiments demonstrate that our method surpasses prior state-of-the-art approaches on multiple segmentation datasets with a remarkable boost in inference speed, while maintaining strong understanding capabilities.
Authors: Dean L Slack, G Thomas Hudson, Thomas Winterbottom, Noura Al Moubayed
Abstract: Inspired by the performance and scalability of autoregressive large language models (LLMs), transformer-based models have seen recent success in the visual domain. This study investigates a transformer adaptation for video prediction with a simple end-to-end approach, comparing various spatiotemporal self-attention layouts. Focusing on causal modeling of physical simulations over time; a common shortcoming of existing video-generative approaches, we attempt to isolate spatiotemporal reasoning via physical object tracking metrics and unsupervised training on physical simulation datasets. We introduce a simple yet effective pure transformer model for autoregressive video prediction, utilizing continuous pixel-space representations for video prediction. Without the need for complex training strategies or latent feature-learning components, our approach significantly extends the time horizon for physically accurate predictions by up to 50% when compared with existing latent-space approaches, while maintaining comparable performance on common video quality metrics. In addition, we conduct interpretability experiments to identify network regions that encode information useful to perform accurate estimations of PDE simulation parameters via probing models, and find that this generalizes to the estimation of out-of-distribution simulation parameters. This work serves as a platform for further attention-based spatiotemporal modeling of videos via a simple, parameter efficient, and interpretable approach.
Authors: Yuhan Liu, Lianhui Qin, Shengjie Wang
Abstract: Large Vision-Language Models (VLMs) have achieved remarkable progress in multimodal understanding, yet they struggle when reasoning over information-intensive images that densely interleave textual annotations with fine-grained graphical elements. The main challenges lie in precisely localizing critical cues in dense layouts and multi-hop reasoning to integrate dispersed evidence. We propose Speculative Verdict (SV), a training-free framework inspired by speculative decoding that combines multiple lightweight draft experts with a large verdict model. In the draft stage, small VLMs act as draft experts to generate reasoning paths that provide diverse localization candidates; in the verdict stage, a strong VLM synthesizes these paths to produce the final answer, minimizing computational cost while recovering correct answers. To further improve efficiency and accuracy, SV introduces a consensus expert selection mechanism that forwards only high-agreement reasoning paths to the verdict. Empirically, SV achieves consistent gains on challenging information-intensive and high-resolution visual question answering benchmarks, including InfographicVQA, ChartMuseum, ChartQAPro, and HR-Bench 4K. By synthesizing correct insights from multiple partially accurate reasoning paths, SV achieves both error correction and cost-efficiency compared to large proprietary models or training pipelines. Code is available at https://github.com/Tinaliu0123/speculative-verdict
Authors: Ritik Shah, Marco F Duarte
Abstract: Hyperspectral sensors capture dense spectra per pixel but suffer from low spatial resolution, causing blurred boundaries and mixed-pixel effects. Co-registered companion sensors such as multispectral, RGB, or panchromatic cameras provide high-resolution spatial detail, motivating hyperspectral super-resolution through the fusion of hyperspectral and multispectral images (HSI-MSI). Existing deep learning based methods achieve strong performance but rely on opaque regressors that lack interpretability and often fail when the MSI has very few bands. We propose SpectraMorph, a physics-guided self-supervised fusion framework with a structured latent space. Instead of direct regression, SpectraMorph enforces an unmixing bottleneck: endmember signatures are extracted from the low-resolution HSI, and a compact multilayer perceptron predicts abundance-like maps from the MSI. Spectra are reconstructed by linear mixing, with training performed in a self-supervised manner via the MSI sensor's spectral response function. SpectraMorph produces interpretable intermediates, trains in under a minute, and remains robust even with a single-band (pan-chromatic) MSI. Experiments on synthetic and real-world datasets show SpectraMorph consistently outperforming state-of-the-art unsupervised/self-supervised baselines while remaining very competitive against supervised baselines.
Authors: Nimrod Berman, Omkar Joglekar, Eitan Kosman, Dotan Di Castro, Omri Azencot
Abstract: Recent advances in generative modeling have positioned diffusion models as state-of-the-art tools for sampling from complex data distributions. While these models have shown remarkable success across single-modality domains such as images and audio, extending their capabilities to Modality Translation (MT), translating information across different sensory modalities, remains an open challenge. Existing approaches often rely on restrictive assumptions, including shared dimensionality, Gaussian source priors, and modality-specific architectures, which limit their generality and theoretical grounding. In this work, we propose the Latent Denoising Diffusion Bridge Model (LDDBM), a general-purpose framework for modality translation based on a latent-variable extension of Denoising Diffusion Bridge Models. By operating in a shared latent space, our method learns a bridge between arbitrary modalities without requiring aligned dimensions. We introduce a contrastive alignment loss to enforce semantic consistency between paired samples and design a domain-agnostic encoder-decoder architecture tailored for noise prediction in latent space. Additionally, we propose a predictive loss to guide training toward accurate cross-domain translation and explore several training strategies to improve stability. Our approach supports arbitrary modality pairs and performs strongly on diverse MT tasks, including multi-view to 3D shape generation, image super-resolution, and multi-view scene synthesis. Comprehensive experiments and ablations validate the effectiveness of our framework, establishing a new strong baseline in general modality translation. For more information, see our project page: https://sites.google.com/view/lddbm/home.
Authors: Guocheng Gordon Qian, Ruihang Zhang, Tsai-Shien Chen, Yusuf Dalva, Anujraaj Argo Goyal, Willi Menapace, Ivan Skorokhodov, Meng Dong, Arpit Sahni, Daniil Ostashev, Ju Hu, Sergey Tulyakov, Kuan-Chieh Jackson Wang
Abstract: Despite their impressive visual fidelity, existing personalized generative models lack interactive control over spatial composition and scale poorly to multiple subjects. To address these limitations, we present LayerComposer, an interactive framework for personalized, multi-subject text-to-image generation. Our approach introduces two main contributions: (1) a layered canvas, a novel representation in which each subject is placed on a distinct layer, enabling occlusion-free composition; and (2) a locking mechanism that preserves selected layers with high fidelity while allowing the remaining layers to adapt flexibly to the surrounding context. Similar to professional image-editing software, the proposed layered canvas allows users to place, resize, or lock input subjects through intuitive layer manipulation. Our versatile locking mechanism requires no architectural changes, relying instead on inherent positional embeddings combined with a new complementary data sampling strategy. Extensive experiments demonstrate that LayerComposer achieves superior spatial control and identity preservation compared to the state-of-the-art methods in multi-subject personalized image generation.
Authors: Yihao Meng, Hao Ouyang, Yue Yu, Qiuyu Wang, Wen Wang, Ka Leong Cheng, Hanlin Wang, Yixuan Li, Cheng Chen, Yanhong Zeng, Yujun Shen, Huamin Qu
Abstract: State-of-the-art text-to-video models excel at generating isolated clips but fall short of creating the coherent, multi-shot narratives, which are the essence of storytelling. We bridge this "narrative gap" with HoloCine, a model that generates entire scenes holistically to ensure global consistency from the first shot to the last. Our architecture achieves precise directorial control through a Window Cross-Attention mechanism that localizes text prompts to specific shots, while a Sparse Inter-Shot Self-Attention pattern (dense within shots but sparse between them) ensures the efficiency required for minute-scale generation. Beyond setting a new state-of-the-art in narrative coherence, HoloCine develops remarkable emergent abilities: a persistent memory for characters and scenes, and an intuitive grasp of cinematic techniques. Our work marks a pivotal shift from clip synthesis towards automated filmmaking, making end-to-end cinematic creation a tangible future. Our code is available at: https://holo-cine.github.io/.
Authors: Trajan Murphy, Akshunna S. Dogra, Hanfeng Gu, Caleb Meredith, Mark Kon, Julio Enrique Castrillion-Candas
Abstract: ''Noisy'' datasets (regimes with low signal to noise ratios, small sample sizes, faulty data collection, etc) remain a key research frontier for classification methods with both theoretical and practical implications. We introduce FINDER, a rigorous framework for analyzing generic classification problems, with tailored algorithms for noisy datasets. FINDER incorporates fundamental stochastic analysis ideas into the feature learning and inference stages to optimally account for the randomness inherent to all empirical datasets. We construct ''stochastic features'' by first viewing empirical datasets as realizations from an underlying random field (without assumptions on its exact distribution) and then mapping them to appropriate Hilbert spaces. The Kosambi-Karhunen-Lo\'eve expansion (KLE) breaks these stochastic features into computable irreducible components, which allow classification over noisy datasets via an eigen-decomposition: data from different classes resides in distinct regions, identified by analyzing the spectrum of the associated operators. We validate FINDER on several challenging, data-deficient scientific domains, producing state of the art breakthroughs in: (i) Alzheimer's Disease stage classification, (ii) Remote sensing detection of deforestation. We end with a discussion on when FINDER is expected to outperform existing methods, its failure modes, and other limitations.
Authors: Drew B. Thomas
Abstract: This paper presents a novel methodology for classifying early modern religious images by using Large Language Models (LLMs) and vector databases in combination with Retrieval-Augmented Generation (RAG). The approach leverages the full-page context of book illustrations from the Holy Roman Empire, allowing the LLM to generate detailed descriptions that incorporate both visual and textual elements. These descriptions are then matched to relevant Iconclass codes through a hybrid vector search. This method achieves 87% and 92% precision at five and four levels of classification, significantly outperforming traditional image and keyword-based searches. By employing full-page descriptions and RAG, the system enhances classification accuracy, offering a powerful tool for large-scale analysis of early modern visual archives. This interdisciplinary approach demonstrates the growing potential of LLMs and RAG in advancing research within art history and digital humanities.
Authors: Adam Diamant
Abstract: This study develops an AI-based pose estimation pipeline to enable precise quantification of movement kinematics in resistance training. Using video data from Wolf et al. (2025), which compared lengthened partial (pROM) and full range-of-motion (fROM) training across eight upper-body exercises in 26 participants, 280 recordings were processed to extract frame-level joint-angle trajectories. After filtering and smoothing, per-set metrics were derived, including range of motion (ROM), tempo, and concentric/eccentric phase durations. A random-effects meta-analytic model was applied to account for within-participant and between-exercise variability. Results show that pROM repetitions were performed with a smaller ROM and shorter overall durations, particularly during the eccentric phase of movement. Variance analyses revealed that participant-level differences, rather than exercise-specific factors, were the primary driver of variation, although there is substantial evidence of heterogeneous treatment effects. We then introduce a novel metric, \%ROM, which is the proportion of full ROM achieved during pROM, and demonstrate that this definition of lengthened partials remains relatively consistent across exercises. Overall, these findings suggest that lengthened partials differ from full ROM training not only in ROM, but also in execution dynamics and consistency, highlighting the potential of AI-based methods for advancing research and improving resistance training prescription.
Authors: Gabriel Y. Arteaga, Marius Aasan, Rwiddhi Chakraborty, Martine Hjelkrem-Tan, Thalles Silva, Michael Kampffmeyer, Ad\'in Ram\'irez Rivera
Abstract: Prototypical self-supervised learning methods consistently suffer from partial prototype collapse, where multiple prototypes converge to nearly identical representations. This undermines their central purpose -- providing diverse and informative targets to guide encoders toward rich representations -- and has led practitioners to over-parameterize prototype sets or add ad-hoc regularizers, which mitigate symptoms rather than address the root cause. We empirically trace the collapse to the joint optimization of encoders and prototypes, which encourages a type of shortcut learning: early in training prototypes drift toward redundant representations that minimize loss without necessarily enhancing representation diversity. To break the joint optimization, we introduce a fully decoupled training strategy that learns prototypes and encoders under separate objectives. Concretely, we model prototypes as a Gaussian mixture updated with an online EM-style procedure, independent of the encoder's loss. This simple yet principled decoupling eliminates prototype collapse without explicit regularization and yields consistently diverse prototypes and stronger downstream performance.
Authors: Rahul Raja, Arpita Vats
Abstract: Question Answering (QA) systems have traditionally relied on structured text data, but the rapid growth of multimedia content (images, audio, video, and structured metadata) has introduced new challenges and opportunities for retrieval-augmented QA. In this survey, we review recent advancements in QA systems that integrate multimedia retrieval pipelines, focusing on architectures that align vision, language, and audio modalities with user queries. We categorize approaches based on retrieval methods, fusion techniques, and answer generation strategies, and analyze benchmark datasets, evaluation protocols, and performance tradeoffs. Furthermore, we highlight key challenges such as cross-modal alignment, latency-accuracy tradeoffs, and semantic grounding, and outline open problems and future research directions for building more robust and context-aware QA systems leveraging multimedia data.
Authors: Mert Bulent Sariyildiz, Philippe Weinzaepfel, Guillaume Bono, Gianluca Monaci, Christian Wolf
Abstract: One key aspect of spatially aware robots is the ability to "find their bearings", ie. to correctly situate themselves in previously seen spaces. In this work, we focus on this particular scenario of continuous robotics operations, where information observed before an actual episode start is exploited to optimize efficiency. We introduce a new model, Kinaema, and agent, capable of integrating a stream of visual observations while moving in a potentially large scene, and upon request, processing a query image and predicting the relative position of the shown space with respect to its current position. Our model does not explicitly store an observation history, therefore does not have hard constraints on context length. It maintains an implicit latent memory, which is updated by a transformer in a recurrent way, compressing the history of sensor readings into a compact representation. We evaluate the impact of this model in a new downstream task we call "Mem-Nav". We show that our large-capacity recurrent model maintains a useful representation of the scene, navigates to goals observed before the actual episode start, and is computationally efficient, in particular compared to classical transformers with attention over an observation history.
Authors: Mahtab Movaheddrad, Laurence Palmer, C. -C. Jay Kuo
Abstract: Image dehazing is a restoration task that aims to recover a clear image from a single hazy input. Traditional approaches rely on statistical priors and the physics-based atmospheric scattering model to reconstruct the haze-free image. While recent state-of-the-art methods are predominantly based on deep learning architectures, these models often involve high computational costs and large parameter sizes, making them unsuitable for resource-constrained devices. In this work, we propose GUSL-Dehaze, a Green U-Shaped Learning approach to image dehazing. Our method integrates a physics-based model with a green learning (GL) framework, offering a lightweight, transparent alternative to conventional deep learning techniques. Unlike neural network-based solutions, GUSL-Dehaze completely avoids deep learning. Instead, we begin with an initial dehazing step using a modified Dark Channel Prior (DCP), which is followed by a green learning pipeline implemented through a U-shaped architecture. This architecture employs unsupervised representation learning for effective feature extraction, together with feature-engineering techniques such as the Relevant Feature Test (RFT) and the Least-Squares Normal Transform (LNT) to maintain a compact model size. Finally, the dehazed image is obtained via a transparent supervised learning strategy. GUSL-Dehaze significantly reduces parameter count while ensuring mathematical interpretability and achieving performance on par with state-of-the-art deep learning models.
Authors: Zixuan Wu, Hengyuan Zhang, Ting-Hsuan Chen, Yuliang Guo, David Paz, Xinyu Huang, Liu Ren
Abstract: Parking is a critical pillar of driving safety. While recent end-to-end (E2E) approaches have achieved promising in-domain results, robustness under domain shifts (e.g., weather and lighting changes) remains a key challenge. Rather than relying on additional data, in this paper, we propose Dino-Diffusion Parking (DDP), a domain-agnostic autonomous parking pipeline that integrates visual foundation models with diffusion-based planning to enable generalized perception and robust motion planning under distribution shifts. We train our pipeline in CARLA at regular setting and transfer it to more adversarial settings in a zero-shot fashion. Our model consistently achieves a parking success rate above 90% across all tested out-of-distribution (OOD) scenarios, with ablation studies confirming that both the network architecture and algorithmic design significantly enhance cross-domain performance over existing baselines. Furthermore, testing in a 3D Gaussian splatting (3DGS) environment reconstructed from a real-world parking lot demonstrates promising sim-to-real transfer.
Authors: Estelle Chigot, Dennis G. Wilson, Meriem Ghrib, Fabrice Jimenez, Thomas Oberlin
Abstract: Deep vision models are now mature enough to be integrated in industrial and possibly critical applications such as autonomous navigation. Yet, data collection and labeling to train such models requires too much efforts and costs for a single company or product. This drawback is more significant in critical applications, where training data must include all possible conditions including rare scenarios. In this perspective, generating synthetic images is an appealing solution, since it allows a cheap yet reliable covering of all the conditions and environments, if the impact of the synthetic-to-real distribution shift is mitigated. In this article, we consider the case of runway detection that is a critical part in autonomous landing systems developed by aircraft manufacturers. We propose an image generation approach based on a commercial flight simulator that complements a few annotated real images. By controlling the image generation and the integration of real and synthetic data, we show that standard object detection models can achieve accurate prediction. We also evaluate their robustness with respect to adverse conditions, in our case nighttime images, that were not represented in the real data, and show the interest of using a customized domain adaptation strategy.
Authors: Tom\'a\v{s} Sou\v{c}ek, Sylvestre-Alvise Rebuffi, Pierre Fernandez, Nikola Jovanovi\'c, Hady Elsahar, Valeriu Lacatusu, Tuan Tran, Alexandre Mourachko
Abstract: Recent years have seen a surge in interest in digital content watermarking techniques, driven by the proliferation of generative models and increased legal pressure. With an ever-growing percentage of AI-generated content available online, watermarking plays an increasingly important role in ensuring content authenticity and attribution at scale. There have been many works assessing the robustness of watermarking to removal attacks, yet, watermark forging, the scenario when a watermark is stolen from genuine content and applied to malicious content, remains underexplored. In this work, we investigate watermark forging in the context of widely used post-hoc image watermarking. Our contributions are as follows. First, we introduce a preference model to assess whether an image is watermarked. The model is trained using a ranking loss on purely procedurally generated images without any need for real watermarks. Second, we demonstrate the model's capability to remove and forge watermarks by optimizing the input image through backpropagation. This technique requires only a single watermarked image and works without knowledge of the watermarking model, making our attack much simpler and more practical than attacks introduced in related work. Third, we evaluate our proposed method on a variety of post-hoc image watermarking models, demonstrating that our approach can effectively forge watermarks, questioning the security of current watermarking approaches. Our code and further resources are publicly available.
Authors: Jan Sobotka, Luca Baroni, J\'an Antol\'ik
Abstract: Decoding visual stimuli from neural population activity is crucial for understanding the brain and for applications in brain-machine interfaces. However, such biological data is often scarce, particularly in primates or humans, where high-throughput recording techniques, such as two-photon imaging, remain challenging or impossible to apply. This, in turn, poses a challenge for deep learning decoding techniques. To overcome this, we introduce MEIcoder, a biologically informed decoding method that leverages neuron-specific most exciting inputs (MEIs), a structural similarity index measure loss, and adversarial training. MEIcoder achieves state-of-the-art performance in reconstructing visual stimuli from single-cell activity in primary visual cortex (V1), especially excelling on small datasets with fewer recorded neurons. Using ablation studies, we demonstrate that MEIs are the main drivers of the performance, and in scaling experiments, we show that MEIcoder can reconstruct high-fidelity natural-looking images from as few as 1,000-2,500 neurons and less than 1,000 training data points. We also propose a unified benchmark with over 160,000 samples to foster future research. Our results demonstrate the feasibility of reliable decoding in early visual system and provide practical insights for neuroscience and neuroengineering applications.
Authors: Shiva Sreeram, Alaa Maalouf, Pratyusha Sharma, Daniela Rus
Abstract: Recently, Sharma et al. suggested a method called Layer-SElective-Rank reduction (LASER) which demonstrated that pruning high-order components of carefully chosen LLM's weight matrices can boost downstream accuracy -- without any gradient-based fine-tuning. Yet LASER's exhaustive, per-matrix search (each requiring full-dataset forward passes) makes it impractical for rapid deployment. We demonstrate that this overhead can be removed and find that: (i) Only a small, carefully chosen subset of matrices needs to be inspected -- eliminating the layer-by-layer sweep, (ii) The gradient of each matrix's singular values pinpoints which matrices merit reduction, (iii) Increasing the factorization search space by allowing matrices rows to cluster around multiple subspaces and then decomposing each cluster separately further reduces overfitting on the original training data and further lifts accuracy by up to 24.6 percentage points, and finally, (iv) we discover that evaluating on just 100 samples rather than the full training data -- both for computing the indicative gradients and for measuring the final accuracy -- suffices to further reduce the search time; we explain that as adaptation to downstream tasks is dominated by prompting style, not dataset size. As a result, we show that combining these findings yields a fast and robust adaptation algorithm for downstream tasks. Overall, with a single gradient step on 100 examples and a quick scan of the top candidate layers and factorization techniques, we can adapt LLMs to new datasets -- entirely without fine-tuning.
Authors: Xueyan Zou, Jianglong Ye, Hao Zhang, Xiaoyu Xiang, Mingyu Ding, Zhaojing Yang, Yong Jae Lee, Zhuowen Tu, Sifei Liu, Xiaolong Wang
Abstract: With the rapid growth of research in AI and robotics now producing over 10,000 papers annually it has become increasingly difficult for researchers to stay up to date. Fast evolving trends, the rise of interdisciplinary work, and the need to explore domains beyond one's expertise all contribute to this challenge. To address these issues, we propose a generalizable pipeline capable of systematically analyzing any research area: identifying emerging trends, uncovering cross domain opportunities, and offering concrete starting points for new inquiry. In this work, we present Real Deep Research (RDR) a comprehensive framework applied to the domains of AI and robotics, with a particular focus on foundation models and robotics advancements. We also briefly extend our analysis to other areas of science. The main paper details the construction of the RDR pipeline, while the appendix provides extensive results across each analyzed topic. We hope this work sheds light for researchers working in the field of AI and beyond.
Authors: Guangqi Jiang, Haoran Chang, Ri-Zhao Qiu, Yutong Liang, Mazeyu Ji, Jiyue Zhu, Zhao Dong, Xueyan Zou, Xiaolong Wang
Abstract: This paper presents GSWorld, a robust, photo-realistic simulator for robotics manipulation that combines 3D Gaussian Splatting with physics engines. Our framework advocates "closing the loop" of developing manipulation policies with reproducible evaluation of policies learned from real-robot data and sim2real policy training without using real robots. To enable photo-realistic rendering of diverse scenes, we propose a new asset format, which we term GSDF (Gaussian Scene Description File), that infuses Gaussian-on-Mesh representation with robot URDF and other objects. With a streamlined reconstruction pipeline, we curate a database of GSDF that contains 3 robot embodiments for single-arm and bimanual manipulation, as well as more than 40 objects. Combining GSDF with physics engines, we demonstrate several immediate interesting applications: (1) learning zero-shot sim2real pixel-to-action manipulation policy with photo-realistic rendering, (2) automated high-quality DAgger data collection for adapting policies to deployment environments, (3) reproducible benchmarking of real-robot manipulation policies in simulation, (4) simulation data collection by virtual teleoperation, and (5) zero-shot sim2real visual reinforcement learning. Website: https://3dgsworld.github.io/.
Authors: Bernd Pfrommer
Abstract: Due to their high temporal resolution and large dynamic range, event cameras are uniquely suited for the analysis of time-periodic signals in an image. In this work we present an efficient and fully asynchronous event camera algorithm for detecting the fundamental frequency at which image pixels flicker. The algorithm employs a second-order digital infinite impulse response (IIR) filter to perform an approximate per-pixel brightness reconstruction and is more robust to high-frequency noise than the baseline method we compare to. We further demonstrate that using the falling edge of the signal leads to more accurate period estimates than the rising edge, and that for certain signals interpolating the zero-level crossings can further increase accuracy. Our experiments find that the outstanding capabilities of the camera in detecting frequencies up to 64kHz for a single pixel do not carry over to full sensor imaging as readout bandwidth limitations become a serious obstacle. This suggests that a hardware implementation closer to the sensor will allow for greatly improved frequency imaging. We discuss the important design parameters for fullsensor frequency imaging and present Frequency Cam, an open-source implementation as a ROS node that can run on a single core of a laptop CPU at more than 50 million events per second. It produces results that are qualitatively very similar to those obtained from the closed source vibration analysis module in Prophesee's Metavision Toolkit. The code for Frequency Cam and a demonstration video can be found at https://github.com/ros-event-camera/frequency_cam
Authors: Deepak Dagar, Dinesh Kumar Vishwakarma
Abstract: Deepfakes, which employ GAN to produce highly realistic facial modification, are widely regarded as the prevailing method. Traditional CNN have been able to identify bogus media, but they struggle to perform well on different datasets and are vulnerable to adversarial attacks due to their lack of robustness. Vision transformers have demonstrated potential in the realm of image classification problems, but they require enough training data. Motivated by these limitations, this publication introduces Tex-ViT (Texture-Vision Transformer), which enhances CNN features by combining ResNet with a vision transformer. The model combines traditional ResNet features with a texture module that operates in parallel on sections of ResNet before each down-sampling operation. The texture module then serves as an input to the dual branch of the cross-attention vision transformer. It specifically focuses on improving the global texture module, which extracts feature map correlation. Empirical analysis reveals that fake images exhibit smooth textures that do not remain consistent over long distances in manipulations. Experiments were performed on different categories of FF++, such as DF, f2f, FS, and NT, together with other types of GAN datasets in cross-domain scenarios. Furthermore, experiments also conducted on FF++, DFDCPreview, and Celeb-DF dataset underwent several post-processing situations, such as blurring, compression, and noise. The model surpassed the most advanced models in terms of generalization, achieving a 98% accuracy in cross-domain scenarios. This demonstrates its ability to learn the shared distinguishing textural characteristics in the manipulated samples. These experiments provide evidence that the proposed model is capable of being applied to various situations and is resistant to many post-processing procedures.
Authors: Ray Congrui Yu, Sherry Wu, Jiang Gui
Abstract: Despite their immense success, deep convolutional neural networks (CNNs) can be difficult to optimize and costly to train due to hundreds of layers within the network depth. Conventional convolutional operations are fundamentally limited by their linear nature along with fixed activations, where many layers are needed to learn meaningful patterns in data. Because of the sheer size of these networks, this approach is simply computationally inefficient, and poses overfitting or gradient explosion risks, especially in small datasets. As a result, we introduce a "plug-in" module, called Residual Kolmogorov-Arnold Network (RKAN). Our module is highly compact, so it can be easily added into any stage (level) of traditional deep networks, where it learns to integrate supportive polynomial feature transformations to existing convolutional frameworks. RKAN offers consistent improvements over baseline models in different vision tasks and widely tested benchmarks, accomplishing cutting-edge performance on them.
Authors: Shrisha Bharadwaj, Haiwen Feng, Giorgio Becherini, Victoria Fernandez Abrevaya, Michael J. Black
Abstract: Manipulating the illumination of a 3D scene within a single image represents a fundamental challenge in computer vision and graphics. This problem has traditionally been addressed using inverse rendering techniques, which involve explicit 3D asset reconstruction and costly ray-tracing simulations. Meanwhile, recent advancements in visual foundation models suggest that a new paradigm could soon be possible -- one that replaces explicit physical models with networks that are trained on large amounts of image and video data. In this paper, we exploit the implicit scene understanding of a video diffusion model, particularly Stable Video Diffusion, to relight a single image. We introduce GenLit, a framework that distills the ability of a graphics engine to perform light manipulation into a video-generation model, enabling users to directly insert and manipulate a point light in the 3D world within a given image and generate results directly as a video sequence. We find that a model fine-tuned on only a small synthetic dataset generalizes to real-world scenes, enabling single-image relighting with plausible and convincing shadows and inter-reflections. Our results highlight the ability of video foundation models to capture rich information about lighting, material, and shape, and our findings indicate that such models, with minimal training, can be used to perform relighting without explicit asset reconstruction or ray-tracing. . Project page: https://genlit.is.tue.mpg.de/.
Authors: Yilei Jiang, Weihong Li, Yiyuan Zhang, Minghong Cai, Xiangyu Yue
Abstract: While Diffusion Models (DM) exhibit remarkable performance across various image generative tasks, they nonetheless reflect the inherent bias presented in the training set. As DMs are now widely used in real-world applications, these biases could perpetuate a distorted worldview and hinder opportunities for minority groups. Existing methods on debiasing DMs usually requires model retraining with a human-crafted reference dataset or additional classifiers, which suffer from two major limitations: (1) collecting reference datasets causes expensive annotation cost; (2) the debiasing performance is heavily constrained by the quality of the reference dataset or the additional classifier. To address the above limitations, we propose FairGen, a plug-and-play method that learns attribute latent directions in a self-discovering manner, thus eliminating the reliance on such reference dataset. Specifically, FairGen consists of two parts: a set of attribute adapters and a distribution indicator. Each adapter in the set aims to learn an attribute latent direction, and is optimized via noise composition through a self-discovering process. Then, the distribution indicator is multiplied by the set of adapters to guide the generation process towards the prescribed distribution. Our method enables debiasing multiple attributes in DMs simultaneously, while remaining lightweight and easily integrable with other DMs, eliminating the need for retraining. Extensive experiments on debiasing gender, racial, and their intersectional biases show that our method outperforms previous SOTA by a large margin.
Authors: Lixiong Qin, Shilong Ou, Miaoxuan Zhang, Jiangning Wei, Yuhang Zhang, Xiaoshuai Song, Yuchen Liu, Mei Wang, Weiran Xu
Abstract: Faces and humans are crucial elements in social interaction and are widely included in everyday photos and videos. Therefore, a deep understanding of faces and humans will enable multi-modal assistants to achieve improved response quality and broadened application scope. Currently, the multi-modal assistant community lacks a comprehensive and scientific evaluation of face and human understanding abilities. In this paper, we first propose a hierarchical ability taxonomy that includes three levels of abilities. Then, based on this taxonomy, we collect images and annotations from publicly available datasets in the face and human community and build a semi-automatic data pipeline to produce problems for the new benchmark. Finally, the obtained Face-Human-Bench includes a development set and a test set, each with 1800 problems, supporting both English and Chinese. We conduct evaluations over 25 mainstream multi-modal large language models (MLLMs) with our Face-Human-Bench, focusing on the correlation between abilities, the impact of the relative position of targets on performance, and the impact of Chain of Thought (CoT) prompting on performance. We also explore which abilities of MLLMs need to be supplemented by specialist models. The dataset and evaluation code have been made publicly available at https://face-human-bench.github.io.
Authors: Yunuo Chen, Junli Cao, Vidit Goel, Sergei Korolev, Chenfanfu Jiang, Jian Ren, Sergey Tulyakov, Anil Kag
Abstract: We present a novel video generation framework that integrates 3-dimensional geometry and dynamic awareness. To achieve this, we augment 2D videos with 3D point trajectories and align them in pixel space. The resulting 3D-aware video dataset, PointVid, is then used to fine-tune a latent diffusion model, enabling it to track 2D objects with 3D Cartesian coordinates. Building on this, we regularize the shape and motion of objects in the video to eliminate undesired artifacts, e.g., non-physical deformation. Consequently, we enhance the quality of generated RGB videos and alleviate common issues like object morphing, which are prevalent in current video models due to a lack of shape awareness. With our 3D augmentation and regularization, our model is capable of handling contact-rich scenarios such as task-oriented videos, where 3D information is essential for perceiving shape and motion of interacting solids. Our method can be seamlessly integrated into existing video diffusion models to improve their visual plausibility.
Authors: Qiwei Wang, Shaoxun Wu, Yujiao Shi
Abstract: This paper addresses the problem of weakly supervised cross-view localization, where the goal is to estimate the pose of a ground camera relative to a satellite image with noisy ground truth annotations. A common approach to bridge the cross-view domain gap for pose estimation is Bird's-Eye View (BEV) synthesis. However, existing methods struggle with height ambiguity due to the lack of depth information in ground images and satellite height maps. Previous solutions either assume a flat ground plane or rely on complex models, such as cross-view transformers. We propose BevSplat, a novel method that resolves height ambiguity by using feature-based Gaussian primitives. Each pixel in the ground image is represented by a 3D Gaussian with semantic and spatial features, which are synthesized into a BEV feature map for relative pose estimation. Additionally, to address challenges with panoramic query images, we introduce an icosphere-based supervision strategy for the Gaussian primitives. We validate our method on the widely used KITTI and VIGOR datasets, which include both pinhole and panoramic query images. Experimental results show that BevSplat significantly improves localization accuracy over prior approaches.
Authors: Xuyang Fang, Sion Hannuna, Neill Campbell, Edwin Simpson
Abstract: Automated livestock monitoring is crucial for precision farming, but robust computer vision models are hindered by a lack of datasets reflecting real-world group challenges. We introduce the 8-Calves dataset, a challenging benchmark for multi-animal detection, tracking, and identification. It features a one-hour video of eight Holstein Friesian calves in a barn, with frequent occlusions, motion blur, and diverse poses. A semi-automated pipeline using a fine-tuned YOLOv8 detector and ByteTrack, followed by manual correction, provides over 537,000 bounding boxes with temporal identity labels. We benchmark 28 object detectors, showing near-perfect performance on a lenient IoU threshold (mAP50: 95.2-98.9%) but significant divergence on stricter metrics (mAP50:95: 56.5-66.4%), highlighting fine-grained localization challenges. Our identification benchmark across 23 models reveals a trade-off: scaling model size improves classification accuracy but compromises retrieval. Smaller architectures like ConvNextV2 Nano achieve the best balance (73.35% accuracy, 50.82% Top-1 KNN). Pre-training focused on semantic learning (e.g., BEiT) yielded superior transferability. For tracking, leading methods achieve high detection accuracy (MOTA > 0.92) but struggle with identity preservation (IDF1 $\approx$ 0.27), underscoring a key challenge in occlusion-heavy scenarios. The 8-Calves dataset bridges a gap by providing temporal richness and realistic challenges, serving as a resource for advancing agricultural vision models. The dataset and code are available at https://huggingface.co/datasets/tonyFang04/8-calves.
Authors: Suraj Singh, Anastasia Batsheva, Oleg Y. Rogov, Ahmed Bouridane
Abstract: Modern image restoration and super-resolution methods utilize deep learning due to its superior performance compared to traditional algorithms. However, deep learning typically requires large training datasets, which are rarely available in astrophotography. Deep Image Prior (DIP) bypasses this constraint by performing blind training on a single image. Although effective in some cases, DIP often suffers from overfitting, artifact generation, and instability. To overcome these issues and improve general performance, this work proposes DIPLI - a framework that shifts from single-frame to multi-frame training using the Back Projection technique, combined with optical flow estimation via the TVNet model, and replaces deterministic predictions with unbiased Monte Carlo estimation obtained through Langevin dynamics. A comprehensive evaluation compares the method against Lucky Imaging, a classical computer vision technique still widely used in astronomical image reconstruction, DIP, the transformer-based model RVRT, and the diffusion-based model DiffIR2VR-Zero. Experiments on synthetic datasets demonstrate consistent improvements, with the method outperforming baselines for SSIM, PSNR, LPIPS, and DISTS metrics in the majority of cases. In addition to superior reconstruction quality, the model also requires far fewer input images than Lucky Imaging and is less prone to overfitting or artifact generation. Evaluation on real-world astronomical data, where domain shifts typically hinder generalization, shows that the method maintains high reconstruction quality, confirming practical robustness.
Authors: Max Gutbrod, David Rauber, Danilo Weber Nunes, Christoph Palm
Abstract: The growing reliance on Artificial Intelligence (AI) in critical domains such as healthcare demands robust mechanisms to ensure the trustworthiness of these systems, especially when faced with unexpected or anomalous inputs. This paper introduces the Open Medical Imaging Benchmarks for Out-Of-Distribution Detection (OpenMIBOOD), a comprehensive framework for evaluating out-of-distribution (OOD) detection methods specifically in medical imaging contexts. OpenMIBOOD includes three benchmarks from diverse medical domains, encompassing 14 datasets divided into covariate-shifted in-distribution, near-OOD, and far-OOD categories. We evaluate 24 post-hoc methods across these benchmarks, providing a standardized reference to advance the development and fair comparison of OOD detection methods. Results reveal that findings from broad-scale OOD benchmarks in natural image domains do not translate to medical applications, underscoring the critical need for such benchmarks in the medical field. By mitigating the risk of exposing AI models to inputs outside their training distribution, OpenMIBOOD aims to support the advancement of reliable and trustworthy AI systems in healthcare. The repository is available at https://github.com/remic-othr/OpenMIBOOD.
Authors: Tzu-Yun Tseng, Alexey Nekrasov, Malcolm Burdorf, Bastian Leibe, Julie Stephany Berrio, Mao Shan, Zhenxing Ming, Stewart Worrall
Abstract: Existing autonomous driving datasets are predominantly oriented towards well-structured urban settings and favourable weather conditions, leaving the complexities of rural environments and adverse weather conditions largely unaddressed. Although some datasets encompass variations in weather and lighting, bad weather scenarios do not appear often. Rainfall can significantly impair sensor functionality, introducing noise and reflections in LiDAR and camera data and reducing the system's capabilities for reliable environmental perception and safe navigation. This paper introduces the Panoptic-CUDAL dataset, a novel dataset purpose-built for panoptic segmentation in rural areas subject to rain. By recording high-resolution LiDAR, camera, and pose data, Panoptic-CUDAL offers a diverse, information-rich dataset in a challenging scenario. We present the analysis of the recorded data and provide baseline results for panoptic, semantic segmentation, and 3D occupancy prediction methods on LiDAR point clouds. The dataset can be found here: https://robotics.sydney.edu.au/our-research/intelligent-transportation-systems, https://vision.rwth-aachen.de/panoptic-cudal
URLs: https://robotics.sydney.edu.au/our-research/intelligent-transportation-systems,, https://vision.rwth-aachen.de/panoptic-cudal
Authors: Linfeng Tang, Yeda Wang, Zhanchuan Cai, Junjun Jiang, Jiayi Ma
Abstract: Current image fusion methods struggle to address the composite degradations encountered in real-world imaging scenarios and lack the flexibility to accommodate user-specific requirements. In response to these challenges, we propose a controllable image fusion framework with language-vision prompts, termed ControlFusion, which adaptively neutralizes composite degradations. On the one hand, we develop a degraded imaging model that integrates physical imaging mechanisms, including the Retinex theory and atmospheric scattering principle, to simulate composite degradations, thereby providing potential for addressing real-world complex degradations from the data level. On the other hand, we devise a prompt-modulated restoration and fusion network that dynamically enhances features with degradation prompts, enabling our method to accommodate composite degradation of varying levels. Specifically, considering individual variations in quality perception of users, we incorporate a text encoder to embed user-specified degradation types and severity levels as degradation prompts. We also design a spatial-frequency collaborative visual adapter that autonomously perceives degradations in source images, thus eliminating the complete dependence on user instructions. Extensive experiments demonstrate that ControlFusion outperforms SOTA fusion methods in fusion quality and degradation handling, particularly in countering real-world and compound degradations with various levels. The source code is publicly available at https://github.com/Linfeng-Tang/ControlFusion.
Authors: Zebin Yao, Lei Ren, Huixing Jiang, Chen Wei, Xiaojie Wang, Ruifan Li, Fangxiang Feng
Abstract: Subject-driven image generation aims to synthesize novel scenes that faithfully preserve subject identity from reference images while adhering to textual guidance. However, existing methods struggle with a critical trade-off between fidelity and efficiency. Tuning-based approaches rely on time-consuming and resource-intensive, subject-specific optimization, while zero-shot methods often fail to maintain adequate subject consistency. In this work, we propose FreeGraftor, a training-free framework that addresses these limitations through cross-image feature grafting. Specifically, FreeGraftor leverages semantic matching and position-constrained attention fusion to transfer visual details from reference subjects to the generated images. Additionally, our framework introduces a novel noise initialization strategy to preserve the geometry priors of reference subjects, facilitating robust feature matching. Extensive qualitative and quantitative experiments demonstrate that our method enables precise subject identity transfer while maintaining text-aligned scene synthesis. Without requiring model fine-tuning or additional training, FreeGraftor significantly outperforms existing zero-shot and training-free approaches in both subject fidelity and text alignment. Furthermore, our framework can seamlessly extend to multi-subject generation, making it practical for real-world deployment. Our code is available at https://github.com/Nihukat/FreeGraftor.
Authors: Daniel Sungho Jung, Kyoung Mu Lee
Abstract: Hands are essential to human interaction, and exploring contact between hands and the world can promote comprehensive understanding of their function. Recently, there have been growing number of hand interaction datasets that cover interaction with object, other hand, scene, and body. Despite the significance of the task and increasing high-quality data, how to effectively learn dense hand contact estimation remains largely underexplored. There are two major challenges for learning dense hand contact estimation. First, there exists class imbalance issue from hand contact datasets where majority of regions are not in contact. Second, hand contact datasets contain spatial imbalance issue with most of hand contact exhibited in finger tips, resulting in challenges for generalization towards contacts in other hand regions. To tackle these issues, we present a framework that learns dense HAnd COntact estimation (HACO) from imbalanced data. To resolve the class imbalance issue, we introduce balanced contact sampling, which builds and samples from multiple sampling groups that fairly represent diverse contact statistics for both contact and non-contact vertices. Moreover, to address the spatial imbalance issue, we propose vertex-level class-balanced (VCB) loss, which incorporates spatially varying contact distribution by separately reweighting loss contribution of each vertex based on its contact frequency across dataset. As a result, we effectively learn to predict dense hand contact estimation with large-scale hand contact data without suffering from class and spatial imbalance issue. The codes are available at https://github.com/dqj5182/HACO_RELEASE.
Authors: Jian Xiao, Zijie Song, Jialong Hu, Hao Cheng, Jia Li, Zhenzhen Hu, Richang Hong
Abstract: Recent progress in text-video retrieval has been largely driven by contrastive learning. However, existing methods often overlook the effect of the modality gap, which causes anchor representations to undergo in-place optimization (i.e., optimization tension) that limits their alignment capacity. Moreover, noisy hard negatives further distort the semantics of anchors. To address these issues, we propose GARE, a Gap-Aware Retrieval framework that introduces a learnable, pair-specific increment $\Delta_{ij}$ between text $t_i$ and video $v_j$, redistributing gradients to relieve optimization tension and absorb noise. We derive $\Delta_{ij}$ via a multivariate first-order Taylor expansion of the InfoNCE loss under a trust-region constraint, showing that it guides updates along locally consistent descent directions. A lightweight neural module conditioned on the semantic gap couples increments across batches for structure-aware correction. Furthermore, we regularize $\Delta$ through a variational information bottleneck with relaxed compression, enhancing stability and semantic consistency. Experiments on four benchmarks demonstrate that GARE consistently improves alignment accuracy and robustness, validating the effectiveness of gap-aware tension mitigation. Code is available at https://github.com/musicman217/GARE-text-video-retrieval.
URLs: https://github.com/musicman217/GARE-text-video-retrieval.
Authors: Die Chen, Zhiwen Li, Cen Chen, Yuexiang Xie, Xiaodan Li, Jinyan Ye, Yingda Chen, Yaliang Li
Abstract: Text-to-image diffusion models have gained widespread application across various domains, demonstrating remarkable creative potential. However, the strong generalization capabilities of diffusion models can inadvertently lead to the generation of not-safe-for-work (NSFW) content, posing significant risks to their safe deployment. While several concept erasure methods have been proposed to mitigate the issue associated with NSFW content, a comprehensive evaluation of their effectiveness across various scenarios remains absent. To bridge this gap, we introduce a full-pipeline toolkit specifically designed for concept erasure and conduct the first systematic study of NSFW concept erasure methods. By examining the interplay between the underlying mechanisms and empirical observations, we provide in-depth insights and practical guidance for the effective application of concept erasure methods in various real-world scenarios, with the aim of advancing the understanding of content safety in diffusion models and establishing a solid foundation for future research and development in this critical area.
Authors: Jian Liu, Jing Xu, Song Guo, Jing Li, Jingfeng Guo, Jiaao Yu, Haohan Weng, Biwen Lei, Xianghui Yang, Zhuo Chen, Fangqi Zhu, Tao Han, Chunchao Guo
Abstract: Existing pretrained models for 3D mesh generation often suffer from data biases and produce low-quality results, while global reinforcement learning (RL) methods rely on object-level rewards that struggle to capture local structure details. To address these challenges, we present Mesh-RFT, a novel fine-grained reinforcement fine-tuning framework that employs Masked Direct Preference Optimization (M-DPO) to enable localized refinement via quality-aware face masking. To facilitate efficient quality evaluation, we introduce an objective topology-aware scoring system to evaluate geometric integrity and topological regularity at both object and face levels through two metrics: Boundary Edge Ratio (BER) and Topology Score (TS). By integrating these metrics into a fine-grained RL strategy, Mesh-RFT becomes the first method to optimize mesh quality at the granularity of individual faces, resolving localized errors while preserving global coherence. Experiment results show that our M-DPO approach reduces Hausdorff Distance (HD) by 24.6% and improves Topology Score (TS) by 3.8% over pre-trained models, while outperforming global DPO methods with a 17.4% HD reduction and 4.9% TS gain. These results demonstrate Mesh-RFT's ability to improve geometric integrity and topological regularity, achieving new state-of-the-art performance in production-ready mesh generation. Project Page: https://hitcslj.github.io/mesh-rft/.
Authors: Xiang Li, Yong Tao, Siyuan Zhang, Siwei Liu, Zhitong Xiong, Chunbo Luo, Lu Liu, Mykola Pechenizkiy, Xiao Xiang Zhu, Tianjin Huang
Abstract: Earth observation foundation models have shown strong generalization across multiple Earth observation tasks, but their robustness under real-world perturbations remains underexplored. To bridge this gap, we introduce REOBench, the first comprehensive benchmark for evaluating the robustness of Earth observation foundation models across six tasks and twelve types of image corruptions, including both appearance-based and geometric perturbations. To ensure realistic and fine-grained evaluation, our benchmark focuses on high-resolution optical remote sensing images, which are widely used in critical applications such as urban planning and disaster response. We conduct a systematic evaluation of a broad range of models trained using masked image modeling, contrastive learning, and vision-language pre-training paradigms. Our results reveal that (1) existing Earth observation foundation models experience significant performance degradation when exposed to input corruptions. (2) The severity of degradation varies across tasks, model architectures, backbone sizes, and types of corruption, with performance drop varying from less than 1% to over 20%. (3) Vision-language models show enhanced robustness, particularly in multimodal tasks. REOBench underscores the vulnerability of current Earth observation foundation models to real-world corruptions and provides actionable insights for developing more robust and reliable models. Code and data are publicly available at https://github.com/lx709/REOBench.
Authors: Hainuo Wang, Qiming Hu, Xiaojie Guo
Abstract: Restoring images degraded by adverse weather remains a significant challenge due to the highly non-uniform and spatially heterogeneous nature of weather-induced artifacts, e.g., fine-grained rain streaks versus widespread haze. Accurately estimating the underlying degradation can intuitively provide restoration models with more targeted and effective guidance, enabling adaptive processing strategies. To this end, we propose a Morton-Order Degradation Estimation Mechanism (MODEM) for adverse weather image restoration. Central to MODEM is the Morton-Order 2D-Selective-Scan Module (MOS2D), which integrates Morton-coded spatial ordering with selective state-space models to capture long-range dependencies while preserving local structural coherence. Complementing MOS2D, we introduce a Dual Degradation Estimation Module (DDEM) that disentangles and estimates both global and local degradation priors. These priors dynamically condition the MOS2D modules, facilitating adaptive and context-aware restoration. Extensive experiments and ablation studies demonstrate that MODEM achieves state-of-the-art results across multiple benchmarks and weather types, highlighting its effectiveness in modeling complex degradation dynamics. Our code will be released at https://github.com/hainuo-wang/MODEM.git.
Authors: Yuetong Fang, Deming Zhou, Ziqing Wang, Hongwei Ren, ZeCui Zeng, Lusong Li, Shibo Zhou, Renjing Xu
Abstract: Spiking Neural Networks promise brain-inspired and energy-efficient computation by transmitting information through binary (0/1) spikes. Yet, their performance still lags behind that of artificial neural networks, often assumed to result from information loss caused by sparse and binary activations. In this work, we challenge this long-standing assumption and reveal a previously overlooked frequency bias: spiking neurons inherently suppress high-frequency components and preferentially propagate low-frequency information. This frequency-domain imbalance, we argue, is the root cause of degraded feature representation in SNNs. Empirically, on Spiking Transformers, adopting Avg-Pooling (low-pass) for token mixing lowers performance to 76.73% on Cifar-100, whereas replacing it with Max-Pool (high-pass) pushes the top-1 accuracy to 79.12%. Accordingly, we introduce Max-Former that restores high-frequency signals through two frequency-enhancing operators: (1) extra Max-Pool in patch embedding, and (2) Depth-Wise Convolution in place of self-attention. Notably, Max-Former attains 82.39% top-1 accuracy on ImageNet using only 63.99M parameters, surpassing Spikformer (74.81%, 66.34M) by +7.58%. Extending our insight beyond transformers, our Max-ResNet-18 achieves state-of-the-art performance on convolution-based benchmarks: 97.17% on CIFAR-10 and 83.06% on CIFAR-100. We hope this simple yet effective solution inspires future research to explore the distinctive nature of spiking neural networks. Code is available: https://github.com/bic-L/MaxFormer.
Authors: Vivek Gopalakrishnan, Neel Dey, Polina Golland
Abstract: Determining the 3D pose of a patient from a limited set of 2D X-ray images is a critical task in interventional settings. While preoperative volumetric imaging (e.g., CT and MRI) provides precise 3D localization and visualization of anatomical targets, these modalities cannot be acquired during procedures, where fast 2D imaging (X-ray) is used instead. To integrate volumetric guidance into intraoperative procedures, we present PolyPose, a simple and robust method for deformable 2D/3D registration. PolyPose parameterizes complex 3D deformation fields as a composition of rigid transforms, leveraging the biological constraint that individual bones do not bend in typical motion. Unlike existing methods that either assume no inter-joint movement or fail outright in this under-determined setting, our polyrigid formulation enforces anatomically plausible priors that respect the piecewise-rigid nature of human movement. This approach eliminates the need for expensive deformation regularizers that require patient- and procedure-specific hyperparameter optimization. Across extensive experiments on diverse datasets from orthopedic surgery and radiotherapy, we show that this strong inductive bias enables PolyPose to successfully align the patient's preoperative volume to as few as two X-rays, thereby providing crucial 3D guidance in challenging sparse-view and limited-angle settings where current registration methods fail. Additional visualizations, tutorials, and code are available at https://polypose.csail.mit.edu.
Authors: Peter Robicheaux, Matvei Popov, Anish Madan, Isaac Robinson, Joseph Nelson, Deva Ramanan, Neehar Peri
Abstract: Vision-language models (VLMs) trained on internet-scale data achieve remarkable zero-shot detection performance on common objects like car, truck, and pedestrian. However, state-of-the-art models still struggle to generalize to out-of-distribution classes, tasks and imaging modalities not typically found in their pre-training. Rather than simply re-training VLMs on more visual data, we argue that one should align VLMs to new concepts with annotation instructions containing a few visual examples and rich textual descriptions. To this end, we introduce Roboflow100-VL, a large-scale collection of 100 multi-modal object detection datasets with diverse concepts not commonly found in VLM pre-training. We evaluate state-of-the-art models on our benchmark in zero-shot, few-shot, semi-supervised, and fully-supervised settings, allowing for comparison across data regimes. Notably, we find that VLMs like GroundingDINO and Qwen2.5-VL achieve less than 2% zero-shot accuracy on challenging medical imaging datasets within Roboflow100-VL, demonstrating the need for few-shot concept alignment. Lastly, we discuss our recent CVPR 2025 Foundational FSOD competition and share insights from the community. Notably, the winning team significantly outperforms our baseline by 17 mAP! Our code and dataset are available at https://github.com/roboflow/rf100-vl and https://universe.roboflow.com/rf100-vl/.
URLs: https://github.com/roboflow/rf100-vl, https://universe.roboflow.com/rf100-vl/.
Authors: Kaiyuan Li, Xiaoyue Chen, Chen Gao, Yong Li, Xinlei Chen
Abstract: Large Vision-Language Models (LVLMs) have shown impressive performance across multi-modal tasks by encoding images into thousands of tokens. However, the large number of image tokens results in significant computational overhead, and the use of dynamic high-resolution inputs further increases this burden. Previous approaches have attempted to reduce the number of image tokens through token pruning, typically by selecting tokens based on attention scores or image token diversity. Through empirical studies, we observe that existing methods often overlook the joint impact of pruning on both the current layer's output (local) and the outputs of subsequent layers (global), leading to suboptimal pruning decisions. To address this challenge, we propose Balanced Token Pruning (BTP), a plug-and-play method for pruning vision tokens. Specifically, our method utilizes a small calibration set to divide the pruning process into multiple stages. In the early stages, our method emphasizes the impact of pruning on subsequent layers, whereas in the deeper stages, the focus shifts toward preserving the consistency of local outputs. Extensive experiments across various LVLMs demonstrate the broad effectiveness of our approach on multiple benchmarks. Our method achieves a 78% compression rate while preserving 96.7% of the original models' performance on average. Our code is available at https://github.com/EmbodiedCity/NeurIPS2025-Balanced-Token-Pruning.
URLs: https://github.com/EmbodiedCity/NeurIPS2025-Balanced-Token-Pruning.
Authors: Yi Ding, Ruqi Zhang
Abstract: Reasoning Vision-Language Models (VLMs) have shown promising performance on complex multimodal tasks. However, they still face significant challenges: they are highly sensitive to reasoning errors, require large volumes of annotated data or accurate verifiers, and struggle to generalize beyond specific domains. To address these limitations, we explore self-correction as a strategy to enhance reasoning VLMs. We first conduct an in-depth analysis of reasoning VLMs' self-correction abilities and identify key gaps. Based on our findings, we introduce Sherlock, a self-correction and self-improvement training framework. Sherlock introduces a trajectory-level self-correction objective, a preference data construction method based on visual perturbation, and a dynamic $\beta$ for preference tuning. Once the model acquires self-correction capabilities using only 20k randomly sampled annotated data, it continues to self-improve without external supervision. Built on the Llama3.2-Vision-11B model, Sherlock achieves remarkable results across eight benchmarks, reaching an average accuracy of 64.1 with direct generation and 65.4 after self-correction. It outperforms LLaVA-CoT (63.2), Mulberry (63.9), and LlamaV-o1 (63.4) while using less than 20% of the annotated data.
Authors: Bowen Chen, Keyan Chen, Mohan Yang, Zhengxia Zou, Zhenwei Shi
Abstract: High-resolution (HR) remote sensing imagery plays a vital role in a wide range of applications, including urban planning and environmental monitoring. However, due to limitations in sensors and data transmission links, the images acquired in practice often suffer from resolution degradation. Remote Sensing Image Super-Resolution (RSISR) aims to reconstruct HR images from low-resolution (LR) inputs, providing a cost-effective and efficient alternative to direct HR image acquisition. Existing RSISR methods primarily focus on low-level characteristics in pixel space, while neglecting the high-level understanding of remote sensing scenes. This may lead to semantically inconsistent artifacts in the reconstructed results. Motivated by this observation, our work aims to explore the role of high-level semantic knowledge in improving RSISR performance. We propose a Semantic-Guided Super-Resolution framework, SeG-SR, which leverages Vision-Language Models (VLMs) to extract semantic knowledge from input images and uses it to guide the super resolution (SR) process. Specifically, we first design a Semantic Feature Extraction Module (SFEM) that utilizes a pretrained VLM to extract semantic knowledge from remote sensing images. Next, we propose a Semantic Localization Module (SLM), which derives a series of semantic guidance from the extracted semantic knowledge. Finally, we develop a Learnable Modulation Module (LMM) that uses semantic guidance to modulate the features extracted by the SR network, effectively incorporating high-level scene understanding into the SR pipeline. We validate the effectiveness and generalizability of SeG-SR through extensive experiments: SeG-SR achieves state-of-the-art performance on three datasets, and consistently improves performance across various SR architectures. Notably, for the x4 SR task on UCMerced dataset, it attained a PSNR of 29.3042 dB and an SSIM of 0.7961.
Authors: Xiao Yu, Yan Fang, Xiaojie Jin, Yao Zhao, Yunchao Wei
Abstract: Audio-visual event parsing plays a crucial role in understanding multimodal video content, but existing methods typically rely on offline processing of entire videos with huge model sizes, limiting their real-time applicability. We introduce Online Audio-Visual Event Parsing (On-AVEP), a novel paradigm for parsing audio, visual, and audio-visual events by sequentially analyzing incoming video streams. The On-AVEP task necessitates models with two key capabilities: (1) Accurate online inference, to effectively distinguish events with unclear and limited context in online settings, and (2) Real-time efficiency, to balance high performance with computational constraints. To cultivate these, we propose the Predictive Future Modeling (PreFM) framework featured by (a) predictive multimodal future modeling to infer and integrate beneficial future audio-visual cues, thereby enhancing contextual understanding and (b) modality-agnostic robust representation along with focal temporal prioritization to improve precision and generalization. Extensive experiments on the UnAV-100 and LLP datasets show PreFM significantly outperforms state-of-the-art methods by a large margin with significantly fewer parameters, offering an insightful approach for real-time multimodal video understanding. Code is available at https://github.com/XiaoYu-1123/PreFM.
Authors: Jianyang Gu, Samuel Stevens, Elizabeth G Campolongo, Matthew J Thompson, Net Zhang, Jiaman Wu, Andrei Kopanev, Zheda Mai, Alexander E. White, James Balhoff, Wasila Dahdul, Daniel Rubenstein, Hilmar Lapp, Tanya Berger-Wolf, Wei-Lun Chao, Yu Su
Abstract: Foundation models trained at scale exhibit remarkable emergent behaviors, learning new capabilities beyond their initial training objectives. We find such emergent behaviors in biological vision models via large-scale contrastive vision-language training. To achieve this, we first curate TreeOfLife-200M, comprising 214 million images of living organisms, the largest and most diverse biological organism image dataset to date. We then train BioCLIP 2 on TreeOfLife-200M to distinguish different species. Despite the narrow training objective, BioCLIP 2 yields extraordinary accuracy when applied to various biological visual tasks such as habitat classification and trait prediction. We identify emergent properties in the learned embedding space of BioCLIP 2. At the inter-species level, the embedding distribution of different species aligns closely with functional and ecological meanings (e.g., beak sizes and habitats). At the intra-species level, instead of being diminished, the intra-species variations (e.g., life stages and sexes) are preserved and better separated in subspaces orthogonal to inter-species distinctions. We provide formal proof and analyses to explain why hierarchical supervision and contrastive objectives encourage these emergent properties. Crucially, our results reveal that these properties become increasingly significant with larger-scale training data, leading to a biologically meaningful embedding space.
Authors: Wenhao Tang, Rong Qin, Heng Fang, Fengtao Zhou, Hao Chen, Xiang Li, Ming-Ming Cheng
Abstract: Pre-trained encoders for offline feature extraction followed by multiple instance learning (MIL) aggregators have become the dominant paradigm in computational pathology (CPath), benefiting cancer diagnosis and prognosis. However, performance limitations arise from the absence of encoder fine-tuning for downstream tasks and disjoint optimization with MIL. While slide-level supervised end-to-end (E2E) learning is an intuitive solution to this issue, it faces challenges such as high computational demands and suboptimal results. These limitations motivate us to revisit E2E learning. We argue that prior work neglects inherent E2E optimization challenges, leading to performance disparities compared to traditional two-stage methods. In this paper, we pioneer the elucidation of optimization challenge caused by sparse-attention MIL and propose a novel MIL called ABMILX. It mitigates this problem through global correlation-based attention refinement and multi-head mechanisms. With the efficient multi-scale random patch sampling strategy, an E2E trained ResNet with ABMILX surpasses SOTA foundation models under the two-stage paradigm across multiple challenging benchmarks, while remaining computationally efficient (<10 RTX3090 hours). We show the potential of E2E learning in CPath and calls for greater research focus in this area. The code is https://github.com/DearCaat/E2E-WSI-ABMILX.
Authors: Xingjian Ran, Yixuan Li, Linning Xu, Mulin Yu, Bo Dai
Abstract: Realistic 3D indoor scene synthesis is vital for embodied AI and digital content creation. It can be naturally divided into two subtasks: object generation and layout generation. While recent generative models have significantly advanced object-level quality and controllability, layout generation remains challenging due to limited datasets. Existing methods either overfit to these datasets or rely on predefined constraints to optimize numerical layout that sacrifice flexibility. As a result, they fail to generate scenes that are both open-vocabulary and aligned with fine-grained user instructions. We introduce DirectLayout, a framework that directly generates numerical 3D layouts from text descriptions using generalizable spatial reasoning of large language models (LLMs). DirectLayout decomposes the generation into three stages: producing a Bird's-Eye View (BEV) layout, lifting it into 3D space, and refining object placements. To enable explicit spatial reasoning and help the model grasp basic principles of object placement, we employ Chain-of-Thought (CoT) Activation based on the 3D-Front dataset. Additionally, we design CoT-Grounded Generative Layout Reward to enhance generalization and spatial planning. During inference, DirectLayout addresses asset-layout mismatches via Iterative Asset-Layout Alignment through in-context learning. Extensive experiments demonstrate that DirectLayout achieves impressive semantic consistency, generalization and physical plausibility.
Authors: Quansong He, Xiangde Min, Kaishen Wang, Tao He
Abstract: Medical image segmentation is a critical task in computer vision, with UNet serving as a milestone architecture. The typical component of UNet family is the skip connection, however, their skip connections face two significant limitations: (1) they lack effective interaction between features at different scales, and (2) they rely on simple concatenation or addition operations, which constrain efficient information integration. While recent improvements to UNet have focused on enhancing encoder and decoder capabilities, these limitations remain overlooked. To overcome these challenges, we propose a novel multi-scale feature fusion method that reimagines the UNet decoding process as solving an initial value problem (IVP), treating skip connections as discrete nodes. By leveraging principles from the linear multistep method, we propose an adaptive ordinary differential equation method to enable effective multi-scale feature fusion. Our approach is independent of the encoder and decoder architectures, making it adaptable to various U-Net-like networks. Experiments on ACDC, KiTS2023, MSD brain tumor, and ISIC2017/2018 skin lesion segmentation datasets demonstrate improved feature utilization, reduced network parameters, and maintained high performance. The code is available at https://github.com/nayutayuki/FuseUNet.
Authors: Xin Yang (College of Biosystems Engineering and Food Science, Zhejiang University, Key Laboratory of Spectroscopy Sensing, Ministry of Agriculture and Rural Affairs), Ruiming Du (Department of Biological and Environmental Engineering, Cornell University), Hanyang Huang (College of Biosystems Engineering and Food Science, Zhejiang University, Key Laboratory of Spectroscopy Sensing, Ministry of Agriculture and Rural Affairs), Jiayang Xie (College of Biosystems Engineering and Food Science, Zhejiang University, Key Laboratory of Spectroscopy Sensing, Ministry of Agriculture and Rural Affairs), Pengyao Xie (College of Biosystems Engineering and Food Science, Zhejiang University, Key Laboratory of Spectroscopy Sensing, Ministry of Agriculture and Rural Affairs), Leisen Fang (College of Biosystems Engineering and Food Science, Zhejiang University, Key Laboratory of Spectroscopy Sensing, Ministry of Agriculture and Rural Affairs), Ziyue Guo (College of Biosystems Engineering and Food Science, Zhejiang University, Key Laboratory of Spectroscopy Sensing, Ministry of Agriculture and Rural Affairs), Nanjun Jiang (Amway), Yu Jiang (Horticulture Section, School of Integrative Plant Science, Cornell AgriTech), Haiyan Cen (College of Biosystems Engineering and Food Science, Zhejiang University, Key Laboratory of Spectroscopy Sensing, Ministry of Agriculture and Rural Affairs)
Abstract: Organ segmentation of plant point clouds is a prerequisite for the high-resolution and accurate extraction of organ-level phenotypic traits. Although the fast development of deep learning has boosted much research on segmentation of plant point clouds, the existing techniques for organ segmentation still face limitations in resolution, segmentation accuracy, and generalizability across various plant species. In this study, we proposed a novel approach called plant segmentation neural radiance fields (PlantSegNeRF), aiming to directly generate high-precision instance point clouds from multi-view RGB image sequences for a wide range of plant species. PlantSegNeRF performed 2D instance segmentation on the multi-view images to generate instance masks for each organ with a corresponding ID. The multi-view instance IDs corresponding to the same plant organ were then matched and refined using a specially designed instance matching module. The instance NeRF was developed to render an implicit scene, containing color, density, semantic and instance information. The implicit scene was ultimately converted into high-precision plant instance point clouds based on the volume density. The results proved that in semantic segmentation of point clouds, PlantSegNeRF outperformed the commonly used methods, demonstrating an average improvement of 16.1%, 18.3%, 17.8%, and 24.2% in precision, recall, F1-score, and IoU compared to the second-best results on structurally complex species. More importantly, PlantSegNeRF exhibited significant advantages in plant point cloud instance segmentation tasks. Across all plant species, it achieved average improvements of 11.7%, 38.2%, 32.2% and 25.3% in mPrec, mRec, mCov, mWCov, respectively. This study extends the organ-level plant phenotyping and provides a high-throughput way to supply high-quality 3D data for the development of large-scale models in plant science.
Authors: Shiting Xiao, Rishabh Kabra, Yuhang Li, Donghyun Lee, Joao Carreira, Priyadarshini Panda
Abstract: The ability to segment objects based on open-ended language prompts remains a critical challenge, requiring models to ground textual semantics into precise spatial masks while handling diverse and unseen categories. We present OpenWorldSAM, a framework that extends the prompt-driven Segment Anything Model v2 (SAM2) to open-vocabulary scenarios by integrating multi-modal embeddings extracted from a lightweight vision-language model (VLM). Our approach is guided by four key principles: i) Unified prompting: OpenWorldSAM supports a diverse range of prompts, including category-level and sentence-level language descriptions, providing a flexible interface for various segmentation tasks. ii) Efficiency: By freezing the pre-trained components of SAM2 and the VLM, we train only 4.5 million parameters on the COCO-stuff dataset, achieving remarkable resource efficiency. iii) Instance Awareness: We enhance the model's spatial understanding through novel positional tie-breaker embeddings and cross-attention layers, enabling effective segmentation of multiple instances. iv) Generalization: OpenWorldSAM exhibits strong zero-shot capabilities, generalizing well on unseen categories and an open vocabulary of concepts without additional training. Extensive experiments demonstrate that OpenWorldSAM achieves state-of-the-art performance in open-vocabulary semantic, instance, and panoptic segmentation across multiple benchmarks. Code is available at https://github.com/GinnyXiao/OpenWorldSAM.
Authors: Chuan Guo, Inwoo Hwang, Jian Wang, Bing Zhou
Abstract: Text-to-motion generation has experienced remarkable progress in recent years. However, current approaches remain limited to synthesizing motion from short or general text prompts, primarily due to dataset constraints. This limitation undermines fine-grained controllability and generalization to unseen prompts. In this paper, we introduce SnapMoGen, a new text-motion dataset featuring high-quality motion capture data paired with accurate, expressive textual annotations. The dataset comprises 20K motion clips totaling 44 hours, accompanied by 122K detailed textual descriptions averaging 48 words per description (vs. 12 words of HumanML3D). Importantly, these motion clips preserve original temporal continuity as they were in long sequences, facilitating research in long-term motion generation and blending. We also improve upon previous generative masked modeling approaches. Our model, MoMask++, transforms motion into multi-scale token sequences that better exploit the token capacity, and learns to generate all tokens using a single generative masked transformer. MoMask++ achieves state-of-the-art performance on both HumanML3D and SnapMoGen benchmarks. Additionally, we demonstrate the ability to process casual user prompts by employing an LLM to reformat inputs to align with the expressivity and narration style of SnapMoGen. Project webpage: https://snap-research.github.io/SnapMoGen/
Authors: Linwei Chen, Lin Gu, Ying Fu
Abstract: Vision Transformers (ViTs) have significantly advanced computer vision, demonstrating strong performance across various tasks. However, the attention mechanism in ViTs makes each layer function as a low-pass filter, and the stacked-layer architecture in existing transformers suffers from frequency vanishing. This leads to the loss of critical details and textures. We propose a novel, circuit-theory-inspired strategy called Frequency-Dynamic Attention Modulation (FDAM), which can be easily plugged into ViTs. FDAM directly modulates the overall frequency response of ViTs and consists of two techniques: Attention Inversion (AttInv) and Frequency Dynamic Scaling (FreqScale). Since circuit theory uses low-pass filters as fundamental elements, we introduce AttInv, a method that generates complementary high-pass filtering by inverting the low-pass filter in the attention matrix, and dynamically combining the two. We further design FreqScale to weight different frequency components for fine-grained adjustments to the target response function. Through feature similarity analysis and effective rank evaluation, we demonstrate that our approach avoids representation collapse, leading to consistent performance improvements across various models, including SegFormer, DeiT, and MaskDINO. These improvements are evident in tasks such as semantic segmentation, object detection, and instance segmentation. Additionally, we apply our method to remote sensing detection, achieving state-of-the-art results in single-scale settings. The code is available at https://github.com/Linwei-Chen/FDAM.
Authors: Ziteng Wang, Siqi Yang, Limeng Qiao, Lin Ma
Abstract: Despite the success of Vision-Language Models (VLMs) like CLIP in aligning vision and language, their proficiency in detailed, fine-grained visual comprehension remains a key challenge. We present CLIP-IN, a novel framework that bolsters CLIP's fine-grained perception through two core innovations. Firstly, we leverage instruction-editing datasets, originally designed for image manipulation, as a unique source of hard negative image-text pairs. Coupled with a symmetric hard negative contrastive loss, this enables the model to effectively distinguish subtle visual-semantic differences. Secondly, CLIP-IN incorporates long descriptive captions, utilizing rotary positional encodings to capture rich semantic context often missed by standard CLIP. Our experiments demonstrate that CLIP-IN achieves substantial gains on the MMVP benchmark and various fine-grained visual recognition tasks, without compromising robust zero-shot performance on broader classification and retrieval tasks. Critically, integrating CLIP-IN's visual representations into Multimodal Large Language Models significantly reduces visual hallucinations and enhances reasoning abilities. This work underscores the considerable potential of synergizing targeted, instruction-based contrastive learning with comprehensive descriptive information to elevate the fine-grained understanding of VLMs.
Authors: Elman Ghazaei, Erchan Aptoula
Abstract: The Earth's surface is constantly changing, and detecting these changes provides valuable insights that benefit various aspects of human society. While traditional change detection methods have been employed to detect changes from bi-temporal images, these approaches typically require expert knowledge for accurate interpretation. To enable broader and more flexible access to change information by non-expert users, the task of Change Detection Visual Question Answering (CDVQA) has been introduced. However, existing CDVQA methods have been developed under the assumption that training and testing datasets share similar distributions. This assumption does not hold in real-world applications, where domain shifts often occur. In this paper, the CDVQA task is revisited with a focus on addressing domain shift. To this end, a new multi-modal and multi-domain dataset, BrightVQA, is introduced to facilitate domain generalization research in CDVQA. Furthermore, a novel state space model, termed Text-Conditioned State Space Model (TCSSM), is proposed. The TCSSM framework is designed to leverage both bi-temporal imagery and geo-disaster-related textual information in an unified manner to extract domain-invariant features across domains. Input-dependent parameters existing in TCSSM are dynamically predicted by using both bi-temporal images and geo-disaster-related description, thereby facilitating the alignment between bi-temporal visual data and the associated textual descriptions. Extensive experiments are conducted to evaluate the proposed method against state-of-the-art models, and superior performance is consistently demonstrated. The code and dataset will be made publicly available upon acceptance at https://github.com/Elman295/TCSSM.
Authors: Jialiang Kang, Han Shu, Wenshuo Li, Yingjie Zhai, Xinghao Chen
Abstract: Speculative decoding is a widely adopted technique for accelerating inference in large language models (LLMs), yet its application to vision-language models (VLMs) remains underexplored, with existing methods achieving only modest speedups (<1.5x). This gap is increasingly significant as multimodal capabilities become central to large-scale models. We hypothesize that large VLMs can effectively filter redundant image information layer by layer without compromising textual comprehension, whereas smaller draft models struggle to do so. To address this, we introduce Vision-Aware Speculative Decoding (ViSpec), a novel framework tailored for VLMs. ViSpec employs a lightweight vision adaptor module to compress image tokens into a compact representation, which is seamlessly integrated into the draft model's attention mechanism while preserving original image positional information. Additionally, we extract a global feature vector for each input image and augment all subsequent text tokens with this feature to enhance multimodal coherence. To overcome the scarcity of multimodal datasets with long assistant responses, we curate a specialized training dataset by repurposing existing datasets and generating extended outputs using the target VLM with modified prompts. Our training strategy mitigates the risk of the draft model exploiting direct access to the target model's hidden states, which could otherwise lead to shortcut learning when training solely on target model outputs. Extensive experiments validate ViSpec, achieving, to our knowledge, the first substantial speedup in VLM speculative decoding. Code is available at https://github.com/KangJialiang/ViSpec.
Authors: Shuqiao Liang, Jian Liu, Renzhang Chen, Quanlong Guan
Abstract: The increasing realism of synthetic images generated by advanced models such as VAEs, GANs, and LDMs poses significant challenges for synthetic image detection. To address this issue, we explore two artifact types introduced during the generation process: (1) latent distribution deviations and (2) decoding-induced smoothing effects, which manifest as inconsistencies in local textures, edges, and color transitions. Leveraging local pixel dependencies (LPD) properties rooted in Markov Random Fields, we reconstruct synthetic images using neighboring pixel information to expose disruptions in texture continuity and edge coherence. Building upon LPD, we propose FerretNet, a lightweight neural network with only 1.1M parameters that delivers efficient and robust synthetic image detection. Extensive experiments demonstrate that FerretNet, trained exclusively on the 4-class ProGAN dataset, achieves an average accuracy of 97.1% on an open-world benchmark comprising 22 generative models. Our code and datasets are publicly available at https://github.com/xigua7105/FerretNet.
Authors: Md Jueal Mia, M. Hadi Amini
Abstract: Vision-Language Models (VLMs) have remarkable abilities in generating multimodal reasoning tasks. However, potential misuse or safety alignment concerns of VLMs have increased significantly due to different categories of attack vectors. Among various attack vectors, recent studies have demonstrated that image-based perturbations are particularly effective in generating harmful outputs. In the literature, many existing techniques have been proposed to jailbreak VLMs, leading to unstable performance and visible perturbations. In this study, we propose Jailbreaking with Loss-guided Image Perturbation (JaiLIP), a jailbreaking attack in the image space that minimizes a joint objective combining the mean squared error (MSE) loss between clean and adversarial image with the models harmful-output loss. We evaluate our proposed method on VLMs using standard toxicity metrics from Perspective API and Detoxify. Experimental results demonstrate that our method generates highly effective and imperceptible adversarial images, outperforming existing methods in producing toxicity. Moreover, we have evaluated our method in the transportation domain to demonstrate the attacks practicality beyond toxic text generation in specific domain. Our findings emphasize the practical challenges of image-based jailbreak attacks and the need for efficient defense mechanisms for VLMs.
Authors: Song Fei, Tian Ye, Lujia Wang, Lei Zhu
Abstract: Universal image restoration (UIR) aims to recover images degraded by unknown mixtures while preserving semantics -- conditions under which discriminative restorers and UNet-based diffusion priors often oversmooth, hallucinate, or drift. We present LucidFlux, a caption-free UIR framework that adapts a large diffusion transformer (Flux.1) without image captions. LucidFlux introduces a lightweight dual-branch conditioner that injects signals from the degraded input and a lightly restored proxy to respectively anchor geometry and suppress artifacts. Then, a timestep- and layer-adaptive modulation schedule is designed to route these cues across the backbone's hierarchy, in order to yield coarse-to-fine and context-aware updates that protect the global structure while recovering texture. After that, to avoid the latency and instability of text prompts or MLLM captions, we enforce caption-free semantic alignment via SigLIP features extracted from the proxy. A scalable curation pipeline further filters large-scale data for structure-rich supervision. Across synthetic and in-the-wild benchmarks, LucidFlux consistently outperforms strong open-source and commercial baselines, and ablation studies verify the necessity of each component. LucidFlux shows that, for large DiTs, when, where, and what to condition on -- rather than adding parameters or relying on text prompts -- is the governing lever for robust and caption-free universal image restoration in the wild.
Authors: Huu Tien Nguyen, Dac Thai Nguyen, The Minh Duc Nguyen, Trung Thanh Nguyen, Thao Nguyen Truong, Huy Hieu Pham, Johan Barthelemy, Minh Quan Tran, Thanh Tam Nguyen, Quoc Viet Hung Nguyen, Quynh Anh Chau, Hong Son Mai, Thanh Trung Nguyen, Phi Le Nguyen
Abstract: Vision-Language Foundation Models (VLMs), trained on large-scale multimodal datasets, have driven significant advances in Artificial Intelligence (AI) by enabling rich cross-modal reasoning. Despite their success in general domains, applying these models to medical imaging remains challenging due to the limited availability of diverse imaging modalities and multilingual clinical data. Most existing medical VLMs are trained on a subset of imaging modalities and focus primarily on high-resource languages, thus limiting their generalizability and clinical utility. To address these limitations, we introduce a novel Vietnamese-language multimodal medical dataset consisting of 2,757 whole-body PET/CT volumes from independent patients and their corresponding full-length clinical reports. This dataset is designed to fill two pressing gaps in medical AI development: (1) the lack of PET/CT imaging data in existing VLMs training corpora, which hinders the development of models capable of handling functional imaging tasks; and (2) the underrepresentation of low-resource languages, particularly the Vietnamese language, in medical vision-language research. To the best of our knowledge, this is the first dataset to provide comprehensive PET/CT-report pairs in Vietnamese. We further introduce a training framework to enhance VLMs' learning, including data augmentation and expert-validated test sets. We conduct comprehensive experiments benchmarking state-of-the-art VLMs on downstream tasks. The experimental results show that incorporating our dataset significantly improves the performance of existing VLMs. We believe this dataset and benchmark will serve as a pivotal step in advancing the development of more robust VLMs for medical imaging, especially for low-resource languages and clinical use in Vietnamese healthcare. The source code is available at https://github.com/AIoT-Lab-BKAI/ViPET-ReportGen.
Authors: Wenhao Li, Qiangchang Wang, Xianjing Meng, Zhibin Wu, Yilong Yin
Abstract: Few-shot learning (FSL) aims to recognize novel concepts from only a few labeled support samples. Recent studies enhance support features by incorporating additional semantic information or designing complex semantic fusion modules. However, they still suffer from hallucinating semantics that contradict the visual evidence due to the lack of grounding in actual instances, resulting in noisy guidance and costly corrections. To address these issues, we propose a novel framework, bridging Vision and Text with LLMs for Few-Shot Learning (VT-FSL), which constructs precise cross-modal prompts conditioned on Large Language Models (LLMs) and support images, seamlessly integrating them through a geometry-aware alignment. It mainly consists of Cross-modal Iterative Prompting (CIP) and Cross-modal Geometric Alignment (CGA). Specifically, the CIP conditions an LLM on both class names and support images to generate precise class descriptions iteratively in a single structured reasoning pass. These descriptions not only enrich the semantic understanding of novel classes but also enable the zero-shot synthesis of semantically consistent images. The descriptions and synthetic images act respectively as complementary textual and visual prompts, providing high-level class semantics and low-level intra-class diversity to compensate for limited support data. Furthermore, the CGA jointly aligns the fused textual, support, and synthetic visual representations by minimizing the kernelized volume of the 3-dimensional parallelotope they span. It captures global and nonlinear relationships among all representations, enabling structured and consistent multimodal integration. The proposed VT-FSL method establishes new state-of-the-art performance across ten diverse benchmarks, including standard, cross-domain, and fine-grained few-shot learning scenarios. Code is available at https://github.com/peacelwh/VT-FSL.
Authors: Seamie Hayes, Ganesh Sistu, Ciar\'an Eising
Abstract: Self-supervised models have recently achieved notable advancements, particularly in the domain of semantic occupancy prediction. These models utilize sophisticated loss computation strategies to compensate for the absence of ground-truth labels. For instance, techniques such as novel view synthesis, cross-view rendering, and depth estimation have been explored to address the issue of semantic and depth ambiguity. However, such techniques typically incur high computational costs and memory usage during the training stage, especially in the case of novel view synthesis. To mitigate these issues, we propose 3D pseudo-ground-truth labels generated by the foundation models Grounded-SAM and Metric3Dv2, and harness temporal information for label densification. Our 3D pseudo-labels can be easily integrated into existing models, which yields substantial performance improvements, with mIoU increasing by 45\%, from 9.73 to 14.09, when implemented into the OccNeRF model. This stands in contrast to earlier advancements in the field, which are often not readily transferable to other architectures. Additionally, we propose a streamlined model, EasyOcc, achieving 13.86 mIoU. This model conducts learning solely from our labels, avoiding complex rendering strategies mentioned previously. Furthermore, our method enables models to attain state-of-the-art performance when evaluated on the full scene without applying the camera mask, with EasyOcc achieving 7.71 mIoU, outperforming the previous best model by 31\%. These findings highlight the critical importance of foundation models, temporal context, and the choice of loss computation space in self-supervised learning for comprehensive scene understanding.
Authors: Zihan Zhou, Shilin Lu, Shuli Leng, Shaocong Zhang, Zhuming Lian, Xinlei Yu, Adams Wai-Kin Kong
Abstract: Drag-based image editing has long suffered from distortions in the target region, largely because the priors of earlier base models, Stable Diffusion, are insufficient to project optimized latents back onto the natural image manifold. With the shift from UNet-based DDPMs to more scalable DiT with flow matching (e.g., SD3.5, FLUX), generative priors have become significantly stronger, enabling advances across diverse editing tasks. However, drag-based editing has yet to benefit from these stronger priors. This work proposes the first framework to effectively harness FLUX's rich prior for drag-based editing, dubbed DragFlow, achieving substantial gains over baselines. We first show that directly applying point-based drag editing to DiTs performs poorly: unlike the highly compressed features of UNets, DiT features are insufficiently structured to provide reliable guidance for point-wise motion supervision. To overcome this limitation, DragFlow introduces a region-based editing paradigm, where affine transformations enable richer and more consistent feature supervision. Additionally, we integrate pretrained open-domain personalization adapters (e.g., IP-Adapter) to enhance subject consistency, while preserving background fidelity through gradient mask-based hard constraints. Multimodal large language models (MLLMs) are further employed to resolve task ambiguities. For evaluation, we curate a novel Region-based Dragging benchmark (ReD Bench) featuring region-level dragging instructions. Extensive experiments on DragBench-DR and ReD Bench show that DragFlow surpasses both point-based and region-based baselines, setting a new state-of-the-art in drag-based image editing. Code and datasets will be publicly available upon publication.
Authors: Dingyi Yao, Xinyao Han, Ruibo Ming, Zhihang Song, Lihui Peng, Jianming Hu, Danya Yao, Yi Zhang
Abstract: Ensuring the reliability of autonomous driving perception systems requires extensive environment-based testing, yet real-world execution is often impractical. Synthetic datasets have therefore emerged as a promising alternative, offering advantages such as cost-effectiveness, bias free labeling, and controllable scenarios. However, the domain gap between synthetic and real-world datasets remains a major obstacle to model generalization. To address this challenge from a data-centric perspective, this paper introduces a profile extraction and discovery framework for characterizing the style profiles underlying both synthetic and real image datasets. We propose Style Embedding Distribution Discrepancy (SEDD) as a novel evaluation metric. Our framework combines Gram matrix-based style extraction with metric learning optimized for intra-class compactness and inter-class separation to extract style embeddings. Furthermore, we establish a benchmark using publicly available datasets. Experiments are conducted on a variety of datasets and sim-to-real methods, and the results show that our method is capable of quantifying the synthetic-to-real gap. This work provides a standardized profiling-based quality control paradigm that enables systematic diagnosis and targeted enhancement of synthetic datasets, advancing future development of data-driven autonomous driving systems.
Authors: Laura Weihl, Stefan H. Bengtson, Nejc Novak, Malte Pedersen
Abstract: Underwater video monitoring is a promising strategy for assessing marine biodiversity, but the vast volume of uneventful footage makes manual inspection highly impractical. In this work, we explore the use of visual anomaly detection (VAD) based on deep neural networks to automatically identify interesting or anomalous events. We introduce AURA, the first multi-annotator benchmark dataset for underwater VAD, and evaluate four VAD models across two marine scenes. We demonstrate the importance of robust frame selection strategies to extract meaningful video segments. Our comparison against multiple annotators reveals that VAD performance of current models varies dramatically and is highly sensitive to both the amount of training data and the variability in visual content that defines "normal" scenes. Our results highlight the value of soft and consensus labels and offer a practical approach for supporting scientific exploration and scalable biodiversity monitoring.
Authors: Theo Di Piazza, Carole Lazarus, Olivier Nempont, Loic Boussel
Abstract: With the growing volume of CT examinations, there is an increasing demand for automated tools such as organ segmentation, abnormality detection, and report generation to support radiologists in managing their clinical workload. Multi-label classification of 3D Chest CT scans remains a critical yet challenging problem due to the complex spatial relationships inherent in volumetric data and the wide variability of abnormalities. Existing methods based on 3D convolutional neural networks struggle to capture long-range dependencies, while Vision Transformers often require extensive pre-training on large-scale, domain-specific datasets to perform competitively. In this work of academic research, we propose a 2.5D alternative by introducing a new graph-based framework that represents 3D CT volumes as structured graphs, where axial slice triplets serve as nodes processed through spectral graph convolution, enabling the model to reason over inter-slice dependencies while maintaining complexity compatible with clinical deployment. Our method, trained and evaluated on 3 datasets from independent institutions, achieves strong cross-dataset generalization, and shows competitive performance compared to state-of-the-art visual encoders. We further conduct comprehensive ablation studies to evaluate the impact of various aggregation strategies, edge-weighting schemes, and graph connectivity patterns. Additionally, we demonstrate the broader applicability of our approach through transfer experiments on automated radiology report generation and abdominal CT data.
Authors: Kedi Ying, Ruiping Liu, Chongyan Chen, Mingzhe Tao, Hao Shi, Kailun Yang, Jiaming Zhang, Rainer Stiefelhagen
Abstract: Walking assistance in extreme or complex environments remains a significant challenge for people with blindness or low vision (BLV), largely due to the lack of a holistic scene understanding. Motivated by the real-world needs of the BLV community, we build mmWalk, a simulated multi-modal dataset that integrates multi-view sensor and accessibility-oriented features for outdoor safe navigation. Our dataset comprises 120 manually controlled, scenario-categorized walking trajectories with 62k synchronized frames. It contains over 559k panoramic images across RGB, depth, and semantic modalities. Furthermore, to emphasize real-world relevance, each trajectory involves outdoor corner cases and accessibility-specific landmarks for BLV users. Additionally, we generate mmWalkVQA, a VQA benchmark with over 69k visual question-answer triplets across 9 categories tailored for safe and informed walking assistance. We evaluate state-of-the-art Vision-Language Models (VLMs) using zero- and few-shot settings and found they struggle with our risk assessment and navigational tasks. We validate our mmWalk-finetuned model on real-world datasets and show the effectiveness of our dataset for advancing multi-modal walking assistance.
Authors: Xiao He, Huangxuan Zhao, Guojia Wan, Wei Zhou, Yanxing Liu, Juhua Liu, Yongchao Xu, Yong Luo, Dacheng Tao, Bo Du
Abstract: Recent medical vision-language models have shown promise on tasks such as VQA, report generation, and anomaly detection. However, most are adapted to structured adult imaging and underperform in fetal ultrasound, which poses challenges of multi-view image reasoning, numerous diseases, and image diversity. To bridge this gap, we introduce FetalMind, a medical AI system tailored to fetal ultrasound for both report generation and diagnosis. Guided by clinical workflow, we propose Salient Epistemic Disentanglement (SED), which injects an expert-curated bipartite graph into the model to decouple view-disease associations and to steer preference selection along clinically faithful steps via reinforcement learning. This design mitigates variability across diseases and heterogeneity across views, reducing learning bottlenecks while aligning the model's inference with obstetric practice. To train FetalMind at scale, we curate FetalSigma-1M dataset, the first large-scale fetal ultrasound report corpus, comprising 20K reports from twelve medical centers, addressing the scarcity of domain data. Extensive experiments show that FetalMind outperforms open- and closed-source baselines across all gestational stages, achieving +14% average gains and +61.2% higher accuracy on critical conditions while remaining efficient, stable, and scalable. Project Page: https://hexiao0275.github.io/FetalMind.
Authors: Yang Li, Aming Wu, Zihao Zhang, Yahong Han
Abstract: In this paper, we focus on Novel Class Discovery for Point Cloud Segmentation (3D-NCD), aiming to learn a model that can segment unlabeled (novel) 3D classes using only the supervision from labeled (base) 3D classes. The key to this task is to setup the exact correlations between the point representations and their base class labels, as well as the representation correlations between the points from base and novel classes. A coarse or statistical correlation learning may lead to the confusion in novel class inference. lf we impose a causal relationship as a strong correlated constraint upon the learning process, the essential point cloud representations that accurately correspond to the classes should be uncovered. To this end, we introduce a structural causal model (SCM) to re-formalize the 3D-NCD problem and propose a new method, i.e., Joint Learning of Causal Representation and Reasoning. Specifically, we first analyze hidden confounders in the base class representations and the causal relationships between the base and novel classes through SCM. We devise a causal representation prototype that eliminates confounders to capture the causal representations of base classes. A graph structure is then used to model the causal relationships between the base classes' causal representation prototypes and the novel class prototypes, enabling causal reasoning from base to novel classes. Extensive experiments and visualization results on 3D and 2D NCD semantic segmentation demonstrate the superiorities of our method.
Authors: Xinmiao Huang, Qisong He, Zhenglin Huang, Boxuan Wang, Zhuoyun Li, Guangliang Cheng, Yi Dong, Xiaowei Huang
Abstract: Spatial reasoning ability is crucial for Vision Language Models (VLMs) to support real-world applications in diverse domains including robotics, augmented reality, and autonomous navigation. Unfortunately, existing benchmarks are inadequate in assessing spatial reasoning ability, especially the \emph{intrinsic-dynamic} spatial reasoning which is a fundamental aspect of human spatial cognition. In this paper, we propose a unified benchmark, \textbf{Spatial-DISE}, based on a cognitively grounded taxonomy that categorizes tasks into four fundamental quadrants: \textbf{I}ntrinsic-\textbf{S}tatic, Intrinsic-\textbf{D}ynamic, \textbf{E}xtrinsic-Static, and Extrinsic-Dynamic spatial reasoning. Moreover, to address the issue of data scarcity, we develop a scalable and automated pipeline to generate diverse and verifiable spatial reasoning questions, resulting in a new \textbf{Spatial-DISE} dataset that includes Spatial-DISE Bench (559 evaluation VQA pairs) and Spatial-DISE-12K (12K+ training VQA pairs). Our comprehensive evaluation across 28 state-of-the-art VLMs reveals that, current VLMs have a large and consistent gap to human competence, especially on multi-step multi-view spatial reasoning. Spatial-DISE offers a robust framework, valuable dataset, and clear direction for future research toward human-like spatial intelligence. Benchmark, dataset, and code will be publicly released.
Authors: Liao Shen, Wentao Jiang, Yiran Zhu, Jiahe Li, Tiezheng Ge, Zhiguo Cao, Bo Zheng
Abstract: Recent advances in image-to-video (I2V) generation have achieved remarkable progress in synthesizing high-quality, temporally coherent videos from static images. Among all the applications of I2V, human-centric video generation includes a large portion. However, existing I2V models encounter difficulties in maintaining identity consistency between the input human image and the generated video, especially when the person in the video exhibits significant expression changes and movements. This issue becomes critical when the human face occupies merely a small fraction of the image. Since humans are highly sensitive to identity variations, this poses a critical yet under-explored challenge in I2V generation. In this paper, we propose Identity-Preserving Reward-guided Optimization (IPRO), a novel video diffusion framework based on reinforcement learning to enhance identity preservation. Instead of introducing auxiliary modules or altering model architectures, our approach introduces a direct and effective tuning algorithm that optimizes diffusion models using a face identity scorer. To improve performance and accelerate convergence, our method backpropagates the reward signal through the last steps of the sampling chain, enabling richer gradient feedback. We also propose a novel facial scoring mechanism that treats faces in ground-truth videos as facial feature pools, providing multi-angle facial information to enhance generalization. A KL-divergence regularization is further incorporated to stabilize training and prevent overfitting to the reward signal. Extensive experiments on Wan 2.2 I2V model and our in-house I2V model demonstrate the effectiveness of our method. Our project and code are available at https://ipro-alimama.github.io/.
Authors: Yunnan Wang, Fan Lu, Kecheng Zheng, Ziyuan Huang, Ziqiang Li, Wenjun Zeng, Xin Jin
Abstract: Multimodal large language models (MLLMs) integrate image features from visual encoders with LLMs, demonstrating advanced comprehension capabilities. However, mainstream MLLMs are solely supervised by the next-token prediction of textual tokens, neglecting critical vision-centric information essential for analytical abilities. To track this dilemma, we introduce VaCo, which optimizes MLLM representations through Vision-Centric activation and Coordination from multiple vision foundation models (VFMs). VaCo introduces visual discriminative alignment to integrate task-aware perceptual features extracted from VFMs, thereby unifying the optimization of both textual and visual outputs in MLLMs. Specifically, we incorporate the learnable Modular Task Queries (MTQs) and Visual Alignment Layers (VALs) into MLLMs, activating specific visual signals under the supervision of diverse VFMs. To coordinate representation conflicts across VFMs, the crafted Token Gateway Mask (TGM) restricts the information flow among multiple groups of MTQs. Extensive experiments demonstrate that VaCo significantly improves the performance of different MLLMs on various benchmarks, showcasing its superior capabilities in visual comprehension.
Authors: Bingyu Li, Feiyu Wang, Da Zhang, Zhiyuan Zhao, Junyu Gao, Xuelong Li
Abstract: Most existing underwater instance segmentation approaches are constrained by close-vocabulary prediction, limiting their ability to recognize novel marine categories. To support evaluation, we introduce \textbf{MARIS} (\underline{Mar}ine Open-Vocabulary \underline{I}nstance \underline{S}egmentation), the first large-scale fine-grained benchmark for underwater Open-Vocabulary (OV) segmentation, featuring a limited set of seen categories and diverse unseen categories. Although OV segmentation has shown promise on natural images, our analysis reveals that transfer to underwater scenes suffers from severe visual degradation (e.g., color attenuation) and semantic misalignment caused by lack underwater class definitions. To address these issues, we propose a unified framework with two complementary components. The Geometric Prior Enhancement Module (\textbf{GPEM}) leverages stable part-level and structural cues to maintain object consistency under degraded visual conditions. The Semantic Alignment Injection Mechanism (\textbf{SAIM}) enriches language embeddings with domain-specific priors, mitigating semantic ambiguity and improving recognition of unseen categories. Experiments show that our framework consistently outperforms existing OV baselines both In-Domain and Cross-Domain setting on MARIS, establishing a strong foundation for future underwater perception research.
Authors: Yeh Keng Hao, Hsu Tzu Wei, Sun Min
Abstract: With the increasing ubiquity of AR/VR devices, the deployment of deep learning models on edge devices has become a critical challenge. These devices require real-time inference, low power consumption, and minimal latency. Many framework designers face the conundrum of balancing efficiency and performance. We design a light framework that adopts an encoder-decoder architecture and introduces several key contributions aimed at improving both efficiency and accuracy. We apply sparse convolution on a ResNet-18 backbone to exploit the inherent sparsity in hand pose images, achieving a 42% end-to-end efficiency improvement. Moreover, we propose our SPLite decoder. This new architecture significantly boosts the decoding process's frame rate by 3.1x on the Raspberry Pi 5, while maintaining accuracy on par. To further optimize performance, we apply quantization-aware training, reducing memory usage while preserving accuracy (PA-MPJPE increases only marginally from 9.0 mm to 9.1 mm on FreiHAND). Overall, our system achieves a 2.98x speed-up on a Raspberry Pi 5 CPU (BCM2712 quad-core Arm A76 processor). Our method is also evaluated on compound benchmark datasets, demonstrating comparable accuracy to state-of-the-art approaches while significantly enhancing computational efficiency.
Authors: Liu Haojie, Gao Suixiang
Abstract: We present HumanCM, a one-step human motion prediction framework built upon consistency models. Instead of relying on multi-step denoising as in diffusion-based methods, HumanCM performs efficient single-step generation by learning a self-consistent mapping between noisy and clean motion states. The framework adopts a Transformer-based spatiotemporal architecture with temporal embeddings to model long-range dependencies and preserve motion coherence. Experiments on Human3.6M and HumanEva-I demonstrate that HumanCM achieves comparable or superior accuracy to state-of-the-art diffusion models while reducing inference steps by up to two orders of magnitude.
Authors: Sanjay Kumar, Tim Brophy, Reenu Mohandas, Eoin Martino Grua, Ganesh Sistu, Valentina Donzella, Ciaran Eising
Abstract: Robust perception in automated driving requires reliable performance under adverse conditions, where sensors may be affected by partial failures or environmental occlusions. Although existing autonomous driving datasets inherently contain sensor noise and environmental variability, very few enable controlled, parameterised, and reproducible degradations across multiple sensing modalities. This gap limits the ability to systematically evaluate how perception and fusion architectures perform under well-defined adverse conditions. To address this limitation, we introduce the Occluded nuScenes Dataset, a novel extension of the widely used nuScenes benchmark. For the camera modality, we release both the full and mini versions with four types of occlusions, two adapted from public implementations and two newly designed. For radar and LiDAR, we provide parameterised occlusion scripts that implement three types of degradations each, enabling flexible and repeatable generation of occluded data. This resource supports consistent, reproducible evaluation of perception models under partial sensor failures and environmental interference. By releasing the first multi-sensor occlusion dataset with controlled and reproducible degradations, we aim to advance research on robust sensor fusion, resilience analysis, and safety-critical perception in automated driving.
Authors: Peiqin Zhuang, Lei Bai, Yichao Wu, Ding Liang, Luping Zhou, Yali Wang, Wanli Ouyang
Abstract: Recently, action recognition has been dominated by transformer-based methods, thanks to their spatiotemporal contextual aggregation capacities. However, despite the significant progress achieved on scene-related datasets, they do not perform well on motion-sensitive datasets due to the lack of elaborate motion modeling designs. Meanwhile, we observe that the widely-used cost volume in traditional action recognition is highly similar to the affinity matrix defined in self-attention, but equipped with powerful motion modeling capacities. In light of this, we propose to integrate those effective motion modeling properties into the existing transformer in a unified and neat way, with the proposal of the Explicit Motion Information Mining module (EMIM). In EMIM, we propose to construct the desirable affinity matrix in a cost volume style, where the set of key candidate tokens is sampled from the query-based neighboring area in the next frame in a sliding-window manner. Then, the constructed affinity matrix is used to aggregate contextual information for appearance modeling and is converted into motion features for motion modeling as well. We validate the motion modeling capacities of our method on four widely-used datasets, and our method performs better than existing state-of-the-art approaches, especially on motion-sensitive datasets, i.e., Something-Something V1 & V2. Our project is available at https://github.com/PeiqinZhuang/EMIM .
Authors: Wenping Jin, Yuyang Tang, Li Zhu, Fei Guo
Abstract: A recent class of hyperspectral anomaly detection methods that can be trained once on background datasets and then universally deployed -- without per-scene retraining or parameter tuning -- has demonstrated remarkable efficiency and robustness. Building upon this paradigm, we focus on the integration of spectral and spatial cues and introduce a novel "Rebellious Student" framework for complementary feature learning. Unlike conventional teacher-student paradigms driven by imitation, our method intentionally trains the spatial branch to diverge from the spectral teacher, thereby learning complementary spatial patterns that the teacher fails to capture. A two-stage learning strategy is adopted: (1) a spectral enhancement network is first trained via reverse distillation to obtain robust background spectral representations; and (2) a spatial network -- the rebellious student -- is subsequently optimized using decorrelation losses that enforce feature orthogonality while maintaining reconstruction fidelity to avoid irrelevant noise. Once trained, the framework enhances both spectral and spatial background features, enabling parameter-free and training-free anomaly detection when paired with conventional detectors. Experiments on the HAD100 benchmark show substantial improvements over several established baselines with modest computational overhead, confirming the effectiveness of the proposed complementary learning paradigm. Our code is publicly available at https://github.com/xjpp2016/FERS.
Authors: Takehiro Aoshima, Yusuke Shinohara, Byeongseon Park
Abstract: Reward-based fine-tuning of video diffusion models is an effective approach to improve the quality of generated videos, as it can fine-tune models without requiring real-world video datasets. However, it can sometimes be limited to specific performances because conventional reward functions are mainly aimed at enhancing the quality across the whole generated video sequence, such as aesthetic appeal and overall consistency. Notably, the temporal consistency of the generated video often suffers when applying previous approaches to image-to-video (I2V) generation tasks. To address this limitation, we propose Video Consistency Distance (VCD), a novel metric designed to enhance temporal consistency, and fine-tune a model with the reward-based fine-tuning framework. To achieve coherent temporal consistency relative to a conditioning image, VCD is defined in the frequency space of video frame features to capture frame information effectively through frequency-domain analysis. Experimental results across multiple I2V datasets demonstrate that fine-tuning a video generation model with VCD significantly enhances temporal consistency without degrading other performance compared to the previous method.
Authors: Zhou Lei, Pan Gang, Wang Jiahao, Sun Di
Abstract: Image Forgery Localization (IFL) is a crucial task in image forensics, aimed at accurately identifying manipulated or tampered regions within an image at the pixel level. Existing methods typically generate a single deterministic localization map, which often lacks the precision and reliability required for high-stakes applications such as forensic analysis and security surveillance. To enhance the credibility of predictions and mitigate the risk of errors, we introduce an advanced Conditional Bernoulli Diffusion Model (CBDiff). Given a forged image, CBDiff generates multiple diverse and plausible localization maps, thereby offering a richer and more comprehensive representation of the forgery distribution. This approach addresses the uncertainty and variability inherent in tampered regions. Furthermore, CBDiff innovatively incorporates Bernoulli noise into the diffusion process to more faithfully reflect the inherent binary and sparse properties of forgery masks. Additionally, CBDiff introduces a Time-Step Cross-Attention (TSCAttention), which is specifically designed to leverage semantic feature guidance with temporal steps to improve manipulation detection. Extensive experiments on eight publicly benchmark datasets demonstrate that CBDiff significantly outperforms existing state-of-the-art methods, highlighting its strong potential for real-world deployment.
Authors: Matthijs Douze, Alexandr Guzhva, Chengqi Deng, Jeff Johnson, Gergely Szilvasy, Pierre-Emmanuel Mazar\'e, Maria Lomeli, Lucas Hosseini, Herv\'e J\'egou
Abstract: Vector databases typically manage large collections of embedding vectors. Currently, AI applications are growing rapidly, and so is the number of embeddings that need to be stored and indexed. The Faiss library is dedicated to vector similarity search, a core functionality of vector databases. Faiss is a toolkit of indexing methods and related primitives used to search, cluster, compress and transform vectors. This paper describes the trade-off space of vector search and the design principles of Faiss in terms of structure, approach to optimization and interfacing. We benchmark key features of the library and discuss a few selected applications to highlight its broad applicability.
Authors: Hanjia Lyu, Ryan Rossi, Xiang Chen, Md Mehrab Tanjim, Stefano Petrangeli, Somdeb Sarkhel, Jiebo Luo
Abstract: Large Language Models (LLMs) have been shown to enhance the effectiveness of enriching item descriptions, thereby improving the accuracy of recommendation systems. However, most existing approaches either rely on text-only prompting or employ basic multimodal strategies that do not fully exploit the complementary information available from both textual and visual modalities. This paper introduces a novel framework, Cross-Reflection Prompting, termed X-Reflect, designed to address these limitations by prompting Multimodal Large Language Models (MLLMs) to explicitly identify and reconcile supportive and conflicting information between text and images. By capturing nuanced insights from both modalities, this approach generates more comprehensive and contextually rich item representations. Extensive experiments conducted on two widely used benchmarks demonstrate that our method outperforms existing prompting baselines in downstream recommendation accuracy. Furthermore, we identify a U-shaped relationship between text-image dissimilarity and recommendation performance, suggesting the benefit of applying multimodal prompting selectively. To support efficient real-time inference, we also introduce X-Reflect-keyword, a lightweight variant that summarizes image content using keywords and replaces the base model with a smaller backbone, achieving nearly 50% reduction in input length while maintaining competitive performance. This work underscores the importance of integrating multimodal information and presents an effective solution for improving item understanding in multimodal recommendation systems.
Authors: Matthias J. Ehrhardt, Subhadip Mukherjee, Hok Shing Wong
Abstract: We address the optimization problem in a data-driven variational reconstruction framework, where the regularizer is parameterized by an input-convex neural network (ICNN). While gradient-based methods are commonly used to solve such problems, they struggle to effectively handle non-smooth problems which often leads to slow convergence. Moreover, the nested structure of the neural network complicates the application of standard non-smooth optimization techniques, such as proximal algorithms. To overcome these challenges, we reformulate the problem and eliminate the network's nested structure. By relating this reformulation to epigraphical projections of the activation functions, we transform the problem into a convex optimization problem that can be efficiently solved using a primal-dual algorithm. We also prove that this reformulation is equivalent to the original variational problem. Through experiments on several imaging tasks, we show that the proposed approach not only outperforms subgradient methods and even accelerated methods in the smooth setting, but also facilitates the training of the regularizer itself.
Authors: Advait Gadhikar, Tom Jacobs, Chao Zhou, Rebekka Burkholz
Abstract: The performance gap between training sparse neural networks from scratch (PaI) and dense-to-sparse training presents a major roadblock for efficient deep learning. According to the Lottery Ticket Hypothesis, PaI hinges on finding a problem specific parameter initialization. As we show, to this end, determining correct parameter signs is sufficient. Yet, they remain elusive to PaI. To address this issue, we propose Sign-In, which employs a dynamic reparameterization that provably induces sign flips. Such sign flips are complementary to the ones that dense-to-sparse training can accomplish, rendering Sign-In as an orthogonal method. While our experiments and theory suggest performance improvements of PaI, they also carve out the main open challenge to close the gap between PaI and dense-to-sparse training.
Authors: Yang Yue, Zhiqi Chen, Rui Lu, Andrew Zhao, Zhaokai Wang, Yang Yue, Shiji Song, Gao Huang
Abstract: Reinforcement Learning with Verifiable Rewards (RLVR) has recently demonstrated notable success in enhancing the reasoning performance of large language models (LLMs), particularly on mathematics and programming tasks. Similar to how traditional RL helps agents explore and learn new strategies, RLVR is believed to enable LLMs to continuously self-improve, thus acquiring novel reasoning abilities beyond those of the corresponding base models. In this study we critically examine the current state of RLVR by systematically probing the reasoning capability boundaries of RLVR-trained LLMs across various model families, RL algorithms, and math, coding, and visual reasoning benchmarks, using pass@k at large k values as the evaluation metric. Surprisingly, we find that the current training setup does not elicit fundamentally new reasoning patterns. While RLVR-trained models outperform their base models at small k (e.g., k = 1), the base models achieve a higher pass@k score when k is large. Coverage and perplexity analyses show that the observed reasoning abilities originate from and are bounded by the base model. Treating the base model as an upper bound, our quantitative analysis shows that six popular RLVR algorithms perform similarly and remain far from optimal in leveraging the potential of the base model. By contrast, we find that distillation can introduce new reasoning patterns from the teacher and genuinely expand the model's reasoning capabilities. Overall, our findings suggest that current RLVR methods have not yet realized the potential of RL to elicit truly novel reasoning abilities in LLMs. This highlights the need for improved RL paradigms, such as continual scaling and multi-turn agent-environment interaction, to unlock this potential.
Authors: Wenyi Xiao, Leilei Gan
Abstract: When applying reinforcement learning--typically through GRPO--to large vision-language model reasoning struggles to effectively scale reasoning length or generates verbose outputs across all tasks with only marginal gains in accuracy. To address this issue, we present FAST-GRPO, a variant of GRPO that dynamically adapts reasoning depth based on question characteristics. Through empirical analysis, we establish the feasibility of fast-slow thinking in LVLMs by investigating how response length and data distribution affect performance. Inspired by these observations, we introduce two complementary metrics to estimate the difficulty of the questions, guiding the model to determine when fast or slow thinking is more appropriate. Next, we incorporate adaptive length-based rewards and difficulty-aware KL divergence into the GRPO algorithm. Experiments across seven reasoning benchmarks demonstrate that FAST achieves state-of-the-art accuracy with over 10\% relative improvement compared to the base model, while reducing token usage by 32.7-67.3\% compared to previous slow-thinking approaches, effectively balancing reasoning length and accuracy.
Authors: Tien Comlekoglu, J. Quetzalcoatl Toledo-Mar\'in, Douglas W. DeSimone, Shayn M. Peirce, Geoffrey Fox, James A. Glazier
Abstract: Mechanistic, multicellular, agent-based models are commonly used to investigate tissue, organ, and organism-scale biology at single-cell resolution. The Cellular-Potts Model (CPM) is a powerful and popular framework for developing and interrogating these models. CPMs become computationally expensive at large space- and time- scales making application and investigation of developed models difficult. Surrogate models may allow for the accelerated evaluation of CPMs of complex biological systems. However, the stochastic nature of these models means each set of parameters may give rise to different model configurations, complicating surrogate model development. In this work, we leverage denoising diffusion probabilistic models to train a generative AI surrogate of a CPM used to investigate in vitro vasculogenesis. We describe the use of an image classifier to learn the characteristics that define unique areas of a 2-dimensional parameter space. We then apply this classifier to aid in surrogate model selection and verification. Our CPM model surrogate generates model configurations 20,000 timesteps ahead of a reference configuration and demonstrates approximately a 22x reduction in computational time as compared to native code execution. Our work represents a step towards the implementation of DDPMs to develop digital twins of stochastic biological systems.
Authors: Jan Hagnberger, Daniel Musekamp, Mathias Niepert
Abstract: Solving time-dependent Partial Differential Equations (PDEs) using a densely discretized spatial domain is a fundamental problem in various scientific and engineering disciplines, including modeling climate phenomena and fluid dynamics. However, performing these computations directly in the physical space often incurs significant computational costs. To address this issue, several neural surrogate models have been developed that operate in a compressed latent space to solve the PDE. While these approaches reduce computational complexity, they often use Transformer-based attention mechanisms to handle irregularly sampled domains, resulting in increased memory consumption. In contrast, convolutional neural networks allow memory-efficient encoding and decoding but are limited to regular discretizations. Motivated by these considerations, we propose CALM-PDE, a model class that efficiently solves arbitrarily discretized PDEs in a compressed latent space. We introduce a novel continuous convolution-based encoder-decoder architecture that uses an epsilon-neighborhood-constrained kernel and learns to apply the convolution operator to adaptive and optimized query points. We demonstrate the effectiveness of CALM-PDE on a diverse set of PDEs with both regularly and irregularly sampled spatial domains. CALM-PDE is competitive with or outperforms existing baseline methods while offering significant improvements in memory and inference time efficiency compared to Transformer-based methods.
Authors: Hao Fang, Changle Zhou, Jiawei Kong, Kuofeng Gao, Bin Chen, Shu-Tao Xia
Abstract: Large Vision-Language Models (LVLMs) are susceptible to hallucinations, where generated responses seem semantically plausible yet exhibit little or no relevance to the input image. Previous studies reveal that this issue primarily stems from LVLMs' over-reliance on language priors while disregarding the visual information during decoding. To alleviate this issue, we introduce a novel Conditional Pointwise Mutual Information (C-PMI) calibrated decoding strategy, which adaptively strengthens the mutual dependency between generated texts and input images to mitigate hallucinations. Unlike existing methods solely focusing on text token sampling, we propose to jointly model the contributions of visual and textual tokens to C-PMI, formulating hallucination mitigation as a bi-level optimization problem aimed at maximizing mutual information. To solve it, we design a token purification mechanism that dynamically regulates the decoding process by sampling text tokens remaining maximally relevant to the given image, while simultaneously refining image tokens most pertinent to the generated response. Extensive experiments across various benchmarks reveal that the proposed method significantly reduces hallucinations in LVLMs while preserving decoding efficiency.
Authors: Declan Kutscher, David M. Chan, Yutong Bai, Trevor Darrell, Ritwik Gupta
Abstract: Sequence models such as transformers require inputs to be represented as one-dimensional sequences. In vision, this typically involves flattening images using a fixed row-major (raster-scan) order. While full self-attention is permutation-equivariant, modern long-sequence transformers increasingly rely on architectural approximations that break this invariance and introduce sensitivity to patch ordering. We show that patch order significantly affects model performance in such settings, with simple alternatives like column-major or Hilbert curves yielding notable accuracy shifts. Motivated by this, we propose REOrder, a two-stage framework for discovering task-optimal patch orderings. First, we derive an information-theoretic prior by evaluating the compressibility of various patch sequences. Then, we learn a policy over permutations by optimizing a Plackett-Luce policy using REINFORCE. This approach enables efficient learning in a combinatorial permutation space. REOrder improves top-1 accuracy over row-major ordering on ImageNet-1K by up to 3.01% and Functional Map of the World by 13.35%.
Authors: Nikola Jovanovi\'c, Ismail Labiad, Tom\'a\v{s} Sou\v{c}ek, Martin Vechev, Pierre Fernandez
Abstract: Watermarking the outputs of generative models has emerged as a promising approach for tracking their provenance. Despite significant interest in autoregressive image generation models and their potential for misuse, no prior work has attempted to watermark their outputs at the token level. In this work, we present the first such approach by adapting language model watermarking techniques to this setting. We identify a key challenge: the lack of reverse cycle-consistency (RCC), wherein re-tokenizing generated image tokens significantly alters the token sequence, effectively erasing the watermark. To address this and to make our method robust to common image transformations, neural compression, and removal attacks, we introduce (i) a custom tokenizer-detokenizer finetuning procedure that improves RCC, and (ii) a complementary watermark synchronization layer. As our experiments demonstrate, our approach enables reliable and robust watermark detection with theoretically grounded p-values. Code and models are available at https://github.com/facebookresearch/wmar.
Authors: Achraf Ait Laydi, Louis Cueff, Mewen Crespo, Yousef El Mourabit, H\'el\`ene Bouvrais
Abstract: Segmenting cytoskeletal filaments in microscopy images is essential for understanding their cellular roles but remains challenging, especially in dense, complex networks and under noisy or low-contrast image conditions. While deep learning has advanced image segmentation, performance often degrades in these adverse scenarios. Additional challenges include the difficulty of obtaining accurate annotations and managing severe class imbalance. We proposed a novel noise-adaptive attention mechanism, extending the Squeeze-and-Excitation (SE) module, to dynamically adjust to varying noise levels. This Adaptive SE (ASE) mechanism is integrated into a U-Net decoder, with residual encoder blocks, forming a lightweight yet powerful model: ASE_Res_U-Net. We also developed a synthetic-dataset strategy and employed tailored loss functions and evaluation metrics to mitigate class imbalance and ensure fair assessment. ASE_Res_U-Net effectively segmented microtubules in both synthetic and real noisy images, outperforming its ablated variants and state-of-the-art curvilinear-structure segmentation methods. It achieved this while using fewer parameters, making it suitable for resource-constrained environments. Importantly, ASE_Res_U-Net generalised well to other curvilinear structures (blood vessels and nerves) under diverse imaging conditions. Availability and implementation: Original microtubule datasets (synthetic and real noisy images) are available on Zenodo (DOIs: 10.5281/zenodo.14696279 and 10.5281/zenodo.15852660). ASE_Res_UNet model will be shared upon publication.
Authors: Haitian Wang, Xinyu Wang, Yiren Wang, Zichen Geng, Xian Zhang, Yu Zhang, Bo Miao
Abstract: Accurate and efficient skin lesion classification on edge devices is critical for accessible dermatological care but remains challenging due to computational, energy, and privacy constraints. We introduce QANA, a novel quantization-aware neuromorphic architecture for incremental skin lesion classification on resource-limited hardware. QANA effectively integrates ghost modules, efficient channel attention, and squeeze-and-excitation blocks for robust feature representation with low-latency and energy-efficient inference. Its quantization-aware head and spike-compatible transformations enable seamless conversion to spiking neural networks (SNNs) and deployment on neuromorphic platforms. Evaluation on the large-scale HAM10000 benchmark and a real-world clinical dataset shows that QANA achieves 91.6% Top-1 accuracy and 82.4% macro F1 on HAM10000, and 90.8%/81.7% on the clinical dataset, significantly outperforming state-of-the-art CNN-to-SNN models under fair comparison. Deployed on BrainChip Akida hardware, QANA achieves 1.5 ms inference latency and 1.7,mJ energy per image, reducing inference latency and energy use by over 94.6%/98.6% compared to GPU-based CNNs surpassing state-of-the-art CNN-to-SNN conversion baselines. These results demonstrate the effectiveness of QANA for accurate, real-time, and privacy-sensitive medical analysis in edge environments.
Authors: Sara Papi, Maike Z\"ufle, Marco Gaido, Beatrice Savoldi, Danni Liu, Ioannis Douros, Luisa Bentivogli, Jan Niehues
Abstract: Recent advances in large language models have catalyzed the development of multimodal LLMs (MLLMs) that integrate text, speech, and vision within unified frameworks. As MLLMs evolve from narrow, monolingual, task-specific systems to general-purpose instruction-following models, a key frontier lies in evaluating their multilingual and multimodal capabilities over both long and short contexts. However, existing benchmarks fall short in evaluating these dimensions jointly: they are often limited to English, mostly focus on one single modality at a time, rely on short-form contexts, or lack human annotations -- hindering comprehensive assessment of model performance across languages, modalities, and task complexity. To address these gaps, we introduce MCIF (Multimodal Crosslingual Instruction Following), the first multilingual human-annotated benchmark based on scientific talks that is designed to evaluate instruction-following in crosslingual, multimodal settings over both short- and long-form inputs. MCIF spans three core modalities -- speech, vision, and text -- and four diverse languages (English, German, Italian, and Chinese), enabling a comprehensive evaluation of MLLMs' abilities to interpret instructions across languages and combine them with multimodal contextual information. MCIF is released under a CC-BY 4.0 license to encourage open research and progress in MLLMs development.
Authors: Xiuyuan Chen, Jian Zhao, Yuchen Yuan, Tianle Zhang, Huilin Zhou, Zheng Zhu, Ping Hu, Linghe Kong, Chi Zhang, Weiran Huang, Xuelong Li
Abstract: Existing safety evaluation methods for large language models (LLMs) suffer from inherent limitations, including evaluator bias and detection failures arising from model homogeneity, which collectively undermine the robustness of risk evaluation processes. This paper seeks to re-examine the risk evaluation paradigm by introducing a theoretical framework that reconstructs the underlying risk concept space. Specifically, we decompose the latent risk concept space into three mutually exclusive subspaces: the explicit risk subspace (encompassing direct violations of safety guidelines), the implicit risk subspace (capturing potential malicious content that requires contextual reasoning for identification), and the non-risk subspace. Furthermore, we propose RADAR, a multi-agent collaborative evaluation framework that leverages multi-round debate mechanisms through four specialized complementary roles and employs dynamic update mechanisms to achieve self-evolution of risk concept distributions. This approach enables comprehensive coverage of both explicit and implicit risks while mitigating evaluator bias. To validate the effectiveness of our framework, we construct an evaluation dataset comprising 800 challenging cases. Extensive experiments on our challenging testset and public benchmarks demonstrate that RADAR significantly outperforms baseline evaluation methods across multiple dimensions, including accuracy, stability, and self-evaluation risk sensitivity. Notably, RADAR achieves a 28.87% improvement in risk identification accuracy compared to the strongest baseline evaluation method.
Authors: Zehao Ni, Yonghao He, Lingfeng Qian, Jilei Mao, Fa Fu, Wei Sui, Hu Su, Junran Peng, Zhipeng Wang, Bin He
Abstract: In the context of imitation learning, visuomotor-based diffusion policy learning is one of the main directions in robotic manipulation. Most of these approaches rely on point clouds as observation inputs and construct scene representations through point clouds feature learning, which enables them to achieve remarkable accuracy. However, the existing literature lacks an in-depth exploration of vision-only solutions that have significant potential. In this paper, we propose a Vision-Only and single-view Diffusion Policy learning method (VO-DP) that leverages pretrained visual foundation models to achieve effective fusion of semantic and geometric features. We utilize intermediate features from VGGT incorporating semantic features from DINOv2 and geometric features from Alternating Attention blocks. Features are fused via cross-attention and spatially compressed with a CNN to form the input to the policy head. Extensive experiments demonstrate that VO-DP not only outperforms the vision-only baseline DP significantly but also exhibits distinct performance trends against the point cloud-based method DP3: in simulation tasks, VO-DP achieves an average success rate of 64.6% on par with DP3 64.0% and far higher than DP 34.8%, while in real-world tasks, it reaches 87.9%, outperforming both DP3 67.5% and DP 11.2% by a notable margin. Further robustness evaluations confirm that VO-DP remains highly stable under varying conditions including color, size, background, and lighting. Lastly, we open-source a training library for robotic manipulation. Built on Accelerate, this library supports multi-machine and multi-GPU parallel training, as well as mixed precision training. It is compatible with visuomotor policies such as DP, DP3 and VO-DP, and also supports the RoboTwin simulator.
Authors: Nilesh Ramgolam, Gustavo Carneiro, Hsiang-Ting Chen
Abstract: This paper addresses the critical data scarcity that hinders the practical deployment of learning to defer (L2D) systems to the population. We introduce a context-aware, semi-supervised framework that uses meta-learning to generate expert-specific embeddings from only a few demonstrations. We demonstrate the efficacy of a dual-purpose mechanism, where these embeddings are used first to generate a large corpus of pseudo-labels for training, and subsequently to enable on-the-fly adaptation to new experts at test-time. The experiment results on three different datasets confirm that a model trained on these synthetic labels rapidly approaches oracle-level performance, validating the data efficiency of our approach. By resolving a key training bottleneck, this work makes adaptive L2D systems more practical and scalable, paving the way for human-AI collaboration in real-world environments. To facilitate reproducibility and address implementation details not covered in the main text, we provide our source code and training configurations at https://github.com/nil123532/learning-to-defer-to-a-population-with-limited-demonstrations.
URLs: https://github.com/nil123532/learning-to-defer-to-a-population-with-limited-demonstrations.
Authors: Jiacheng Liu, Xinyu Wang, Yuqi Lin, Zhikai Wang, Peiru Wang, Peiliang Cai, Qinming Zhou, Zhengan Yan, Zexuan Yan, Zhengyi Shi, Chang Zou, Yue Ma, Linfeng Zhang
Abstract: Diffusion Models have become a cornerstone of modern generative AI for their exceptional generation quality and controllability. However, their inherent \textit{multi-step iterations} and \textit{complex backbone networks} lead to prohibitive computational overhead and generation latency, forming a major bottleneck for real-time applications. Although existing acceleration techniques have made progress, they still face challenges such as limited applicability, high training costs, or quality degradation. Against this backdrop, \textbf{Diffusion Caching} offers a promising training-free, architecture-agnostic, and efficient inference paradigm. Its core mechanism identifies and reuses intrinsic computational redundancies in the diffusion process. By enabling feature-level cross-step reuse and inter-layer scheduling, it reduces computation without modifying model parameters. This paper systematically reviews the theoretical foundations and evolution of Diffusion Caching and proposes a unified framework for its classification and analysis. Through comparative analysis of representative methods, we show that Diffusion Caching evolves from \textit{static reuse} to \textit{dynamic prediction}. This trend enhances caching flexibility across diverse tasks and enables integration with other acceleration techniques such as sampling optimization and model distillation, paving the way for a unified, efficient inference framework for future multimodal and interactive applications. We argue that this paradigm will become a key enabler of real-time and efficient generative AI, injecting new vitality into both theory and practice of \textit{Efficient Generative Intelligence}.