Authors: L\'eore Bensabath, Mathis Petrovich, G\"ul Varol
Abstract: Our goal is to train a generative model of 3D hand motions, conditioned on natural language descriptions specifying motion characteristics such as handshapes, locations, finger/hand/arm movements. To this end, we automatically build pairs of 3D hand motions and their associated textual labels with unprecedented scale. Specifically, we leverage a large-scale sign language video dataset, along with noisy pseudo-annotated sign categories, which we translate into hand motion descriptions via an LLM that utilizes a dictionary of sign attributes, as well as our complementary motion-script cues. This data enables training a text-conditioned hand motion diffusion model HandMDM, that is robust across domains such as unseen sign categories from the same sign language, but also signs from another sign language and non-sign hand movements. We contribute extensive experimental investigation of these scenarios and will make our trained models and data publicly available to support future research in this relatively new field.
Authors: Kaining Li, Shuwei He, Zihan Xu
Abstract: Human action recognition in long-term videos, characterized by complex backgrounds and subtle action differences, poses significant challenges for traditional deep learning models due to computational overhead, difficulty in capturing long-range temporal dependencies, and limited semantic understanding. While Large Language Models (LLMs) and Large Vision-Language Models (LVLMs) have shown remarkable capabilities in multi-modal understanding and reasoning, their direct application to continuous video streams for fine-grained action recognition remains an open problem. This paper introduces VT-LVLM-AR (Video-Temporal Large Vision-Language Model Adapter for Action Recognition), a novel framework designed to bridge this gap. VT-LVLM-AR comprises a Video-to-Event Mapper (VTEM) that efficiently transforms raw video into compact, semantically rich, and temporally coherent "visual event sequences" through lightweight spatio-temporal feature extraction, adaptive temporal pooling, and conceptual quantization with an event coherence bias. These visual event sequences are then fed into an LVLM-based Action Reasoning module, specifically a frozen LLaVA-1.5 model, adapted using parameter-efficient Prompt Tuning (P-Tuning v2) for action classification. Comprehensive evaluations on the NTU RGB+D and NTU RGB+D 120 datasets demonstrate that VT-LVLM-AR consistently achieves state-of-the-art performance, surpassing existing methods (e.g., 94.1% accuracy on NTU RGB+D X-Sub). Ablation studies confirm the critical contributions of VTEM's components and the efficacy of Prompt Tuning, while human evaluations underscore the interpretability of our visual event representations. This work highlights the immense potential of leveraging LVLMs for robust and interpretable video action understanding through effective video-to-language translation and efficient model adaptation.
Authors: Dexuan He, Xiao Zhou, Wenbin Guan, Liyuan Zhang, Xiaoman Zhang, Sinuo Xu, Ge Wang, Lifeng Wang, Xiaojun Yuan, Xin Sun, Yanfeng Wang, Kun Sun, Ya Zhang, Weidi Xie
Abstract: Rare cancers comprise 20-25% of all malignancies but face major diagnostic challenges due to limited expert availability-especially in pediatric oncology, where they represent over 70% of cases. While pathology vision-language (VL) foundation models show promising zero-shot capabilities for common cancer subtyping, their clinical performance for rare cancers remains limited. Existing multi-instance learning (MIL) methods rely only on visual features, overlooking cross-modal knowledge and compromising interpretability critical for rare cancer diagnosis. To address this limitation, we propose PathPT, a novel framework that fully exploits the potential of vision-language pathology foundation models through spatially-aware visual aggregation and task-specific prompt tuning. Unlike conventional MIL, PathPT converts WSI-level supervision into fine-grained tile-level guidance by leveraging the zero-shot capabilities of VL models, thereby preserving localization on cancerous regions and enabling cross-modal reasoning through prompts aligned with histopathological semantics. We benchmark PathPT on eight rare cancer datasets(four adult and four pediatric) spanning 56 subtypes and 2,910 WSIs, as well as three common cancer datasets, evaluating four state-of-the-art VL models and four MIL frameworks under three few-shot settings. Results show that PathPT consistently delivers superior performance, achieving substantial gains in subtyping accuracy and cancerous region grounding ability. This work advances AI-assisted diagnosis for rare cancers, offering a scalable solution for improving subtyping accuracy in settings with limited access to specialized expertise.
Authors: Jiahao Li, Jiancheng Pan, Yuze Sun, Xiaomeng Huang
Abstract: Ship detection in remote sensing imagery is a critical task with wide-ranging applications, such as maritime activity monitoring, shipping logistics, and environmental studies. However, existing methods often struggle to capture fine-grained semantic information, limiting their effectiveness in complex scenarios. To address these challenges, we propose a novel detection framework that combines Vision-Language Models (VLMs) with a multi-scale adaptive sliding window strategy. To facilitate Semantic-Aware Ship Detection (SASD), we introduce ShipSem-VL, a specialized Vision-Language dataset designed to capture fine-grained ship attributes. We evaluate our framework through three well-defined tasks, providing a comprehensive analysis of its performance and demonstrating its effectiveness in advancing SASD from multiple perspectives.
Authors: Jiawen Lyu, Manu Ramesh, Madison Simonds, Jacquelyn P. Boerman, Amy R. Reibman
Abstract: Few automated video systems are described in the open literature that enable hands-free cataloging and identification (ID) of cows in a dairy herd. In this work, we describe our system, composed of an AutoCattloger, which builds a Cattlog of dairy cows in a herd with a single input video clip per cow, an eidetic cow recognizer which uses no deep learning to ID cows, and a CowFinder, which IDs cows in a continuous stream of video. We demonstrate its value in finding individuals in unlabeled, unsegmented videos of cows walking unconstrained through the holding area of a milking parlor.
Authors: Angela Zhu, Christian Lange, Max Hamilton
Abstract: Species distribution models encode spatial patterns of species occurrence making them effective priors for vision-based species classification when location information is available. In this study, we evaluate various SINR (Spatial Implicit Neural Representations) models as a geographical prior for visual classification of species from iNaturalist observations. We explore the impact of different model configurations and adjust how we handle predictions for species not included in Geo Prior training. Our analysis reveals factors that contribute to the effectiveness of these models as Geo Priors, factors that may differ from making accurate range maps.
Authors: Mahmoud Khalil, Ahmad Khalil, Alioune Ngom
Abstract: Deep learning vision models are typically tailored for specific modalities and often rely on domain-specific assumptions, such as the grid structures used by nearly all existing vision models. In this work, we propose a self-supervised model based on Transformers, which we call Adaptive Superpixel Coding (ASC). The key insight of our model is to overcome the limitations of traditional Vision Transformers, which depend on fixed-size and non-adaptive patch partitioning. Instead, ASC employs adaptive superpixel layers that dynamically adjust to the underlying image content. We analyze key properties of the approach that make it effective, and find that our method outperforms widely-used alternatives on standard image downstream task benchmarks.
Authors: Zhenhao Guo, Rachit Saluja, Tianyuan Yao, Quan Liu, Yuankai Huo, Benjamin Liechty, David J. Pisapia, Kenji Ikemura, Mert R. Sabuncu, Yihe Yang, Ruining Deng
Abstract: Vision-language models (VLMs) have shown considerable potential in digital pathology, yet their effectiveness remains limited for fine-grained, disease-specific classification tasks such as distinguishing between glomerular subtypes. The subtle morphological variations among these subtypes, combined with the difficulty of aligning visual patterns with precise clinical terminology, make automated diagnosis in renal pathology particularly challenging. In this work, we explore how large pretrained VLMs can be effectively adapted to perform fine-grained glomerular classification, even in scenarios where only a small number of labeled examples are available. In this work, we introduce Glo-VLMs, a systematic framework designed to explore the adaptation of VLMs to fine-grained glomerular classification in data-constrained settings. Our approach leverages curated pathology images alongside clinical text prompts to facilitate joint image-text representation learning for nuanced renal pathology subtypes. By assessing various VLMs architectures and adaptation strategies under a few-shot learning paradigm, we explore how both the choice of method and the amount of labeled data impact model performance in clinically relevant scenarios. To ensure a fair comparison, we evaluate all models using standardized multi-class metrics, aiming to clarify the practical requirements and potential of large pretrained models for specialized clinical research applications. As a result, fine-tuning the VLMs achieved 0.7416 accuracy, 0.9045 macro-AUC, and 0.5277 F1-score with only 8 shots per class, demonstrating that even with highly limited supervision, foundation models can be effectively adapted for fine-grained medical image classification.
Authors: Minh-Tan Pham
Abstract: This manuscript presents a series of my selected contributions to the topic of label-efficient learning in computer vision and remote sensing. The central focus of this research is to develop and adapt methods that can learn effectively from limited or partially annotated data, and can leverage abundant unlabeled data in real-world applications. The contributions span both methodological developments and domain-specific adaptations, in particular addressing challenges unique to Earth observation data such as multi-modality, spatial resolution variability, and scene heterogeneity. The manuscript is organized around four main axes including (1) weakly supervised learning for object discovery and detection based on anomaly-aware representations learned from large amounts of background images; (2) multi-task learning that jointly trains on multiple datasets with disjoint annotations to improve performance on object detection and semantic segmentation; (3) self-supervised and supervised contrastive learning with multimodal data to enhance scene classification in remote sensing; and (4) few-shot learning for hierarchical scene classification using both explicit and implicit modeling of class hierarchies. These contributions are supported by extensive experimental results across natural and remote sensing datasets, reflecting the outcomes of several collaborative research projects. The manuscript concludes by outlining ongoing and future research directions focused on scaling and enhancing label-efficient learning for real-world applications.
Authors: Ousmane Youme, Jean Marie Demb\'el\'e, Eugene C. Ezin, Christophe Cambier
Abstract: Convolutional neural networks (CNN) have been used efficiently in several fields, including environmental challenges. In fact, CNN can help with the monitoring of marine litter, which has become a worldwide problem. UAVs have higher resolution and are more adaptable in local areas than satellite images, making it easier to find and count trash. Since the sand is heterogeneous, a basic CNN model encounters plenty of inferences caused by reflections of sand color, human footsteps, shadows, algae present, dunes, holes, and tire tracks. For these types of images, other CNN models, such as CNN-based segmentation methods, may be more appropriate. In this paper, we use an instance-based segmentation method and a panoptic segmentation method that show good accuracy with just a few samples. The model is more robust and less
Authors: Jerry Cao-Xue, Tien Comlekoglu, Keyi Xue, Guanliang Wang, Jiang Li, Gordon Laurie
Abstract: The development of multi-label deep learning models for retinal disease classification is often hindered by the scarcity of large, expertly annotated clinical datasets due to patient privacy concerns and high costs. The recent release of SynFundus-1M, a high-fidelity synthetic dataset with over one million fundus images, presents a novel opportunity to overcome these barriers. To establish a foundational performance benchmark for this new resource, we developed an end-to-end deep learning pipeline, training six modern architectures (ConvNeXtV2, SwinV2, ViT, ResNet, EfficientNetV2, and the RETFound foundation model) to classify eleven retinal diseases using a 5-fold multi-label stratified cross-validation strategy. We further developed a meta-ensemble model by stacking the out-of-fold predictions with an XGBoost classifier. Our final ensemble model achieved the highest performance on the internal validation set, with a macro-average Area Under the Receiver Operating Characteristic Curve (AUC) of 0.9973. Critically, the models demonstrated strong generalization to three diverse, real-world clinical datasets, achieving an AUC of 0.7972 on a combined DR dataset, an AUC of 0.9126 on the AIROGS glaucoma dataset and a macro-AUC of 0.8800 on the multi-label RFMiD dataset. This work provides a robust baseline for future research on large-scale synthetic datasets and establishes that models trained exclusively on synthetic data can accurately classify multiple pathologies and generalize effectively to real clinical images, offering a viable pathway to accelerate the development of comprehensive AI systems in ophthalmology.
Authors: Mohamed Ilyes Lakhal, Richard Bowden
Abstract: The diversity of sign representation is essential for Sign Language Production (SLP) as it captures variations in appearance, facial expressions, and hand movements. However, existing SLP models are often unable to capture diversity while preserving visual quality and modelling non-manual attributes such as emotions. To address this problem, we propose a novel approach that leverages Latent Diffusion Model (LDM) to synthesise photorealistic digital avatars from a generated reference image. We propose a novel sign feature aggregation module that explicitly models the non-manual features (\textit{e.g.}, the face) and the manual features (\textit{e.g.}, the hands). We show that our proposed module ensures the preservation of linguistic content while seamlessly using reference images with different ethnic backgrounds to ensure diversity. Experiments on the YouTube-SL-25 sign language dataset show that our pipeline achieves superior visual quality compared to state-of-the-art methods, with significant improvements on perceptual metrics.
Authors: Aykut Sirma, Angelos Plastropoulos, Argyrios Zolotas, Gilbert Tang
Abstract: Recent advancements in computer vision and deep learning have enhanced disaster-response capabilities, particularly in the rapid assessment of earthquake-affected urban environments. Timely identification of accessible entry points and structural obstacles is essential for effective search-and-rescue (SAR) operations. To address this need, we introduce DRespNeT, a high-resolution dataset specifically developed for aerial instance segmentation of post-earthquake structural environments. Unlike existing datasets, which rely heavily on satellite imagery or coarse semantic labeling, DRespNeT provides detailed polygon-level instance segmentation annotations derived from high-definition (1080p) aerial footage captured in disaster zones, including the 2023 Turkiye earthquake and other impacted regions. The dataset comprises 28 operationally critical classes, including structurally compromised buildings, access points such as doors, windows, and gaps, multiple debris levels, rescue personnel, vehicles, and civilian visibility. A distinctive feature of DRespNeT is its fine-grained annotation detail, enabling differentiation between accessible and obstructed areas, thereby improving operational planning and response efficiency. Performance evaluations using YOLO-based instance segmentation models, specifically YOLOv8-seg, demonstrate significant gains in real-time situational awareness and decision-making. Our optimized YOLOv8-DRN model achieves 92.7% mAP50 with an inference speed of 27 FPS on an RTX-4090 GPU for multi-target detection, meeting real-time operational requirements. The dataset and models support SAR teams and robotic systems, providing a foundation for enhancing human-robot collaboration, streamlining emergency response, and improving survivor outcomes.
Authors: Floris Erich, Naoya Chiba, Abdullah Mustafa, Ryo Hanai, Noriaki Ando, Yusuke Yoshiyasu, Yukiyasu Domae
Abstract: How can we extract complete geometric models of objects that we encounter in our daily life, without having access to commercial 3D scanners? In this paper we present an automated system for generating geometric models of objects from two or more videos. Our system requires the specification of one known point in at least one frame of each video, which can be automatically determined using a fiducial marker such as a checkerboard or Augmented Reality (AR) marker. The remaining frames are automatically positioned in world space by using Structure-from-Motion techniques. By using multiple videos and merging results, a complete object mesh can be generated, without having to rely on hole filling. Code for our system is available from https://github.com/FlorisE/NeuralMeshing.
Authors: Jinyue Song, Hansol Ku, Jayneel Vora, Nelson Lee, Ahmad Kamari, Prasant Mohapatra, Parth Pathak
Abstract: Automotive FMCW radars remain reliable in rain and glare, yet their sparse, noisy point clouds constrain 3-D object detection. We therefore release CoVeRaP, a 21 k-frame cooperative dataset that time-aligns radar, camera, and GPS streams from multiple vehicles across diverse manoeuvres. Built on this data, we propose a unified cooperative-perception framework with middle- and late-fusion options. Its baseline network employs a multi-branch PointNet-style encoder enhanced with self-attention to fuse spatial, Doppler, and intensity cues into a common latent space, which a decoder converts into 3-D bounding boxes and per-point depth confidence. Experiments show that middle fusion with intensity encoding boosts mean Average Precision by up to 9x at IoU 0.9 and consistently outperforms single-vehicle baselines. CoVeRaP thus establishes the first reproducible benchmark for multi-vehicle FMCW-radar perception and demonstrates that affordable radar sharing markedly improves detection robustness. Dataset and code are publicly available to encourage further research.
Authors: Cory Gardner, Byungseok Min, Tae-Hyuk Ahn
Abstract: Anomaly detection and localization in industrial images are essential for automated quality inspection. PaDiM, a prominent method, models the distribution of normal image features extracted by pre-trained Convolutional Neural Networks (CNNs) but reduces dimensionality through random channel selection, potentially discarding structured information. We propose Wavelet-Enhanced PaDiM (WE-PaDiM), which integrates Discrete Wavelet Transform (DWT) analysis with multi-layer CNN features in a structured manner. WE-PaDiM applies 2D DWT to feature maps from multiple backbone layers, selects specific frequency subbands (e.g., LL, LH, HL), spatially aligns them, and concatenates them channel-wise before modeling with PaDiM's multivariate Gaussian framework. This DWT-before-concatenation strategy provides a principled method for feature selection based on frequency content relevant to anomalies, leveraging multi-scale wavelet information as an alternative to random selection. We evaluate WE-PaDiM on the challenging MVTec AD dataset with multiple backbones (ResNet-18 and EfficientNet B0-B6). The method achieves strong performance in anomaly detection and localization, yielding average results of 99.32% Image-AUC and 92.10% Pixel-AUC across 15 categories with per-class optimized configurations. Our analysis shows that wavelet choices affect performance trade-offs: simpler wavelets (e.g., Haar) with detail subbands (HL or LH/HL/HH) often enhance localization, while approximation bands (LL) improve image-level detection. WE-PaDiM thus offers a competitive and interpretable alternative to random feature selection in PaDiM, achieving robust results suitable for industrial inspection with comparable efficiency.
Authors: Zhaoyi Yan, Binghui Chen, Yunfan Liu, Qixiang Ye
Abstract: Knowledge distillation (KD) aims to transfer knowledge from a large-scale teacher model to a lightweight one, significantly reducing computational and storage requirements. However, the inherent learning capacity gap between the teacher and student often hinders the sufficient transfer of knowledge, motivating numerous studies to address this challenge. Inspired by the progressive approximation principle in the Stone-Weierstrass theorem, we propose Expandable Residual Approximation (ERA), a novel KD method that decomposes the approximation of residual knowledge into multiple steps, reducing the difficulty of mimicking the teacher's representation through a divide-and-conquer approach. Specifically, ERA employs a Multi-Branched Residual Network (MBRNet) to implement this residual knowledge decomposition. Additionally, a Teacher Weight Integration (TWI) strategy is introduced to mitigate the capacity disparity by reusing the teacher's head weights. Extensive experiments show that ERA improves the Top-1 accuracy on the ImageNet classification benchmark by 1.41% and the AP on the MS COCO object detection benchmark by 1.40, as well as achieving leading performance across computer vision tasks. Codes and models are available at https://github.com/Zhaoyi-Yan/ERA.
Authors: Ziqi Li, Abderraouf Amrani, Shri Rai, Hamid Laga
Abstract: Reconstructing the 3D geometry, pose, and motion of animals is a long-standing problem, which has a wide range of applications, from biology, livestock management, and animal conservation and welfare to content creation in digital entertainment and Virtual/Augmented Reality (VR/AR). Traditionally, 3D models of real animals are obtained using 3D scanners. These, however, are intrusive, often prohibitively expensive, and difficult to deploy in the natural environment of the animals. In recent years, we have seen a significant surge in deep learning-based techniques that enable the 3D reconstruction, in a non-intrusive manner, of the shape and motion of dynamic objects just from their RGB image and/or video observations. Several papers have explored their application and extension to various types of animals. This paper surveys the latest developments in this emerging and growing field of research. It categorizes and discusses the state-of-the-art methods based on their input modalities, the way the 3D geometry and motion of animals are represented, the type of reconstruction techniques they use, and the training mechanisms they adopt. It also analyzes the performance of some key methods, discusses their strengths and limitations, and identifies current challenges and directions for future research.
Authors: Qifeng Liu, Dawei Zhao, Yabo Dong, Linzhi Shang, Liang Xiao, Juan Wang, Kunkong Zhao, Dongming Lu, Qi Zhu
Abstract: Recent advances in point cloud object detection have increasingly adopted Transformer-based and State Space Models (SSMs), demonstrating strong performance. However, voxelbased representations in these models require strict consistency in input and output dimensions due to their serialized processing, which limits the spatial diffusion capability typically offered by convolutional operations. This limitation significantly affects detection accuracy. Inspired by CNN-based object detection architectures, we propose a novel Voxel Diffusion Module (VDM) to enhance voxel-level representation and diffusion in point cloud data. VDM is composed of sparse 3D convolutions, submanifold sparse convolutions, and residual connections. To ensure computational efficiency, the output feature maps are downsampled to one-fourth of the original input resolution. VDM serves two primary functions: (1) diffusing foreground voxel features through sparse 3D convolutions to enrich spatial context, and (2) aggregating fine-grained spatial information to strengthen voxelwise feature representation. The enhanced voxel features produced by VDM can be seamlessly integrated into mainstream Transformer- or SSM-based detection models for accurate object classification and localization, highlighting the generalizability of our method. We evaluate VDM on several benchmark datasets by embedding it into both Transformerbased and SSM-based models. Experimental results show that our approach consistently improves detection accuracy over baseline models. Specifically, VDM-SSMs achieve 74.7 mAPH (L2) on Waymo, 72.9 NDS on nuScenes, 42.3 mAP on Argoverse 2, and 67.6 mAP on ONCE, setting new stateof-the-art performance across all datasets. Our code will be made publicly available.
Authors: Xiangde Luo, Xiyue Wang, Feyisope Eweje, Xiaoming Zhang, Sen Yang, Ryan Quinton, Jinxi Xiang, Yuchen Li, Yuanfeng Ji, Zhe Li, Yijiang Chen, Colin Bergstrom, Ted Kim, Francesca Maria Olguin, Kelley Yuan, Matthew Abikenari, Andrew Heider, Sierra Willens, Sanjeeth Rajaram, Robert West, Joel Neal, Maximilian Diehn, Ruijiang Li
Abstract: Histopathology is essential for disease diagnosis and treatment decision-making. Recent advances in artificial intelligence (AI) have enabled the development of pathology foundation models that learn rich visual representations from large-scale whole-slide images (WSIs). However, existing models are often trained on disparate datasets using varying strategies, leading to inconsistent performance and limited generalizability. Here, we introduce ELF (Ensemble Learning of Foundation models), a novel framework that integrates five state-of-the-art pathology foundation models to generate unified slide-level representations. Trained on 53,699 WSIs spanning 20 anatomical sites, ELF leverages ensemble learning to capture complementary information from diverse models while maintaining high data efficiency. Unlike traditional tile-level models, ELF's slide-level architecture is particularly advantageous in clinical contexts where data are limited, such as therapeutic response prediction. We evaluated ELF across a wide range of clinical applications, including disease classification, biomarker detection, and response prediction to major anticancer therapies, cytotoxic chemotherapy, targeted therapy, and immunotherapy, across multiple cancer types. ELF consistently outperformed all constituent foundation models and existing slide-level models, demonstrating superior accuracy and robustness. Our results highlight the power of ensemble learning for pathology foundation models and suggest ELF as a scalable and generalizable solution for advancing AI-assisted precision oncology.
Authors: Sun Weikai, Song Shijie, Chi Wenjie
Abstract: Although diffusion model has made good progress in the field of image generation, GAN\cite{huang2023adaptive} still has a large development space due to its unique advantages, such as WGAN\cite{liu2021comparing}, SSGAN\cite{guibas2021adaptive} \cite{zhang2022vsa} \cite{zhou2024adapt} and so on. In this paper, we propose a novel two-flow feedback multi-scale progressive generative adversarial network (MSPG-SEN) for GAN models. This paper has four contributions: 1) : We propose a two-flow feedback multi-scale progressive Generative Adversarial network (MSPG-SEN), which not only improves image quality and human visual perception on the basis of retaining the advantages of the existing GAN model, but also simplifies the training process and reduces the training cost of GAN networks. Our experimental results show that, MSPG-SEN has achieved state-of-the-art generation results on the following five datasets,INKK The dataset is 89.7\%,AWUN The dataset is 78.3\%,IONJ The dataset is 85.5\%,POKL The dataset is 88.7\%,OPIN The dataset is 96.4\%. 2) : We propose an adaptive perception-behavioral feedback loop (APFL), which effectively improves the robustness and training stability of the model and reduces the training cost. 3) : We propose a globally connected two-flow dynamic residual network(). After ablation experiments, it can effectively improve the training efficiency and greatly improve the generalization ability, with stronger flexibility. 4) : We propose a new dynamic embedded attention mechanism (DEMA). After experiments, the attention can be extended to a variety of image processing tasks, which can effectively capture global-local information, improve feature separation capability and feature expression capabilities, and requires minimal computing resources only 88.7\% with INJK With strong cross-task capability.
Authors: Savvas Karatsiolis, Andreas Kamilaris
Abstract: We propose Domain Adaptation via Feature Refinement (DAFR2), a simple yet effective framework for unsupervised domain adaptation under distribution shift. The proposed method synergistically combines three key components: adaptation of Batch Normalization statistics using unlabeled target data, feature distillation from a source-trained model and hypothesis transfer. By aligning feature distributions at the statistical and representational levels, DAFR2 produces robust and domain-invariant feature spaces that generalize across similar domains without requiring target labels, complex architectures or sophisticated training objectives. Extensive experiments on benchmark datasets, including CIFAR10-C, CIFAR100-C, MNIST-C and PatchCamelyon-C, demonstrate that the proposed algorithm outperforms prior methods in robustness to corruption. Theoretical and empirical analyses further reveal that our method achieves improved feature alignment, increased mutual information between the domains and reduced sensitivity to input perturbations.
Authors: Hao Tang, Rongxi Yi, Lei Li, Kaiyi Cao, Jiapeng Zhao, Yihan Xiao, Minghai Shi, Peng Yuan, Yan Xi, Hui Tang, Wei Li, Zhan Wu, Yixin Zhou
Abstract: Conventional computed tomography (CT) lacks the ability to capture dynamic, weight-bearing joint motion. Functional evaluation, particularly after surgical intervention, requires four-dimensional (4D) imaging, but current methods are limited by excessive radiation exposure or incomplete spatial information from 2D techniques. We propose an integrated 4D joint analysis platform that combines: (1) a dual robotic arm cone-beam CT (CBCT) system with a programmable, gantry-free trajectory optimized for upright scanning; (2) a hybrid imaging pipeline that fuses static 3D CBCT with dynamic 2D X-rays using deep learning-based preprocessing, 3D-2D projection, and iterative optimization; and (3) a clinically validated framework for quantitative kinematic assessment. In simulation studies, the method achieved sub-voxel accuracy (0.235 mm) with a 99.18 percent success rate, outperforming conventional and state-of-the-art registration approaches. Clinical evaluation further demonstrated accurate quantification of tibial plateau motion and medial-lateral variance in post-total knee arthroplasty (TKA) patients. This 4D CBCT platform enables fast, accurate, and low-dose dynamic joint imaging, offering new opportunities for biomechanical research, precision diagnostics, and personalized orthopedic care.
Authors: Jincheng Li, Danyang Dong, Menglin Zheng, Jingbo Zhang, Yueqin Hang, Lichi Zhang, Lili Zhao
Abstract: Automatic detection of abnormal cervical cells from Thinprep Cytologic Test (TCT) images is a critical component in the development of intelligent computer-aided diagnostic systems. However, existing algorithms typically fail to effectively model the correlations of visual features, while these spatial correlation features actually contain critical diagnostic information. Furthermore, no detection algorithm has the ability to integrate inter-correlation features of cells with intra-discriminative features of cells, lacking a fusion strategy for the end-to-end detection model. In this work, we propose a hypergraph-based cell detection network that effectively fuses different types of features, combining spatial correlation features and deep discriminative features. Specifically, we use a Multi-level Fusion Sub-network (MLF-SNet) to enhance feature extractioncapabilities. Then we introduce a Cross-level Feature Fusion Strategy with Hypergraph Computation module (CLFFS-HC), to integrate mixed features. Finally, we conducted experiments on three publicly available datasets, and the results demonstrate that our method significantly improves the performance of cervical abnormal cell detection.
Authors: Pi-Wei Chen, Jerry Chun-Wei Lin, Wei-Han Chen, Jia Ji, Zih-Ching Chen, Feng-Hao Yeh, Chao-Chun Chen
Abstract: Pre-trained Vision-Language Models (VLMs) have recently shown promise in detecting anomalies. However, previous approaches are fundamentally limited by their reliance on human-designed prompts and the lack of accessible anomaly samples, leading to significant gaps in context-specific anomaly understanding. In this paper, we propose \textbf{A}daptive \textbf{P}rompt \textbf{T}uning with semantic alignment for anomaly detection (APT), a groundbreaking prior knowledge-free, few-shot framework and overcomes the limitations of traditional prompt-based approaches. APT uses self-generated anomaly samples with noise perturbations to train learnable prompts that capture context-dependent anomalies in different scenarios. To prevent overfitting to synthetic noise, we propose a Self-Optimizing Meta-prompt Guiding Scheme (SMGS) that iteratively aligns the prompts with general anomaly semantics while incorporating diverse synthetic anomaly. Our system not only advances pixel-wise anomaly detection, but also achieves state-of-the-art performance on multiple benchmark datasets without requiring prior knowledge for prompt crafting, establishing a robust and versatile solution for real-world anomaly detection.
Authors: Haodong He, Yancheng Bai, Rui Lan, Xu Duan, Lei Sun, Xiangxiang Chu, Gui-Song Xia
Abstract: The rich textual information of large vision-language models (VLMs) combined with the powerful generative prior of pre-trained text-to-image (T2I) diffusion models has achieved impressive performance in single-image super-resolution (SISR). However, existing methods still face significant challenges in generating clear and accurate regional details, particularly in scenarios involving multiple objects. This challenge primarily stems from a lack of fine-grained regional descriptions and the models' insufficient ability to capture complex prompts. To address these limitations, we propose a Regional Attention Guided Super-Resolution (RAGSR) method that explicitly extracts localized fine-grained information and effectively encodes it through a novel regional attention mechanism, enabling both enhanced detail and overall visually coherent SR results. Specifically, RAGSR localizes object regions in an image and assigns fine-grained caption to each region, which are formatted as region-text pairs as textual priors for T2I models. A regional guided attention is then leveraged to ensure that each region-text pair is properly considered in the attention process while preventing unwanted interactions between unrelated region-text pairs. By leveraging this attention mechanism, our approach offers finer control over the integration of text and image information, thereby effectively overcoming limitations faced by traditional SISR techniques. Experimental results on benchmark datasets demonstrate that our approach exhibits superior performance in generating perceptually authentic visual details while maintaining contextual consistency compared to existing approaches.
Authors: Jiaqi Ma, Guo-Sen Xie, Fang Zhao, Zechao Li
Abstract: Meta-learning aims to uniformly sample homogeneous support-query pairs, characterized by the same categories and similar attributes, and extract useful inductive biases through identical network architectures. However, this identical network design results in over-semantic homogenization. To address this, we propose a novel homologous but heterogeneous network. By treating support-query pairs as dual perspectives, we introduce heterogeneous visual aggregation (HA) modules to enhance complementarity while preserving semantic commonality. To further reduce semantic noise and amplify the uniqueness of heterogeneous semantics, we design a heterogeneous transfer (HT) module. Finally, we propose heterogeneous CLIP (HC) textual information to enhance the generalization capability of multimodal models. In the weakly-supervised few-shot semantic segmentation (WFSS) task, with only 1/24 of the parameters of existing state-of-the-art models, TLG achieves a 13.2\% improvement on Pascal-5\textsuperscript{i} and a 9.7\% improvement on COCO-20\textsuperscript{i}. To the best of our knowledge, TLG is also the first weakly supervised (image-level) model that outperforms fully supervised (pixel-level) models under the same backbone architectures. The code is available at https://github.com/jarch-ma/TLG.
Authors: Mohammad Mohammadzadeh Kalati, Farhad Maleki, Ian McQuillan
Abstract: Predicting and tracking objects in real-world scenarios is a critical challenge in Video Object Segmentation (VOS) tasks. Unsupervised VOS (UVOS) has the additional challenge of finding an initial segmentation of salient objects, which affects the entire process and keeps a permanent uncertainty about the object proposals. Moreover, deformation and fast motion can lead to temporal inconsistencies. To address these problems, we propose Frequent Temporally Integrated Objects (FTIO), a post-processing framework with two key components. First, we introduce a combined criterion to improve object selection, mitigating failures common in UVOS--particularly when objects are small or structurally complex--by extracting frequently appearing salient objects. Second, we present a three-stage method to correct temporal inconsistencies by integrating missing object mask regions. Experimental results demonstrate that FTIO achieves state-of-the-art performance in multi-object UVOS. Code is available at: https://github.com/MohammadMohammadzadehKalati/FTIO
Authors: Yicheng Ji, Jun Zhang, Heming Xia, Jinpeng Chen, Lidan Shou, Gang Chen, Huan Li
Abstract: Video large language models (Vid-LLMs) have shown strong capabilities in understanding video content. However, their reliance on dense video token representations introduces substantial memory and computational overhead in both prefilling and decoding. To mitigate the information loss of recent video token reduction methods and accelerate the decoding stage of Vid-LLMs losslessly, we introduce SpecVLM, a training-free speculative decoding (SD) framework tailored for Vid-LLMs that incorporates staged video token pruning. Building on our novel finding that the draft model's speculation exhibits low sensitivity to video token pruning, SpecVLM prunes up to 90% of video tokens, enabling efficient speculation without sacrificing accuracy. To achieve this, it performs a two-stage pruning process: Stage I selects highly informative tokens guided by attention signals from the verifier (target model), while Stage II prunes remaining redundant ones in a spatially uniform manner. Extensive experiments on four video understanding benchmarks demonstrate the effectiveness and robustness of SpecVLM, which achieves up to 2.68$\times$ decoding speedup for LLaVA-OneVision-72B and 2.11$\times$ speedup for Qwen2.5-VL-32B.
Authors: Thinesh Thiyakesan Ponbagavathi, Kunyu Peng, Alina Roitberg
Abstract: Changes of camera perspective are a common obstacle in driver monitoring. While deep learning and pretrained foundation models show strong potential for improved generalization via lightweight adaptation of the final layers ('probing'), their robustness to unseen viewpoints remains underexplored. We study this challenge by adapting image foundation models to driver monitoring using a single training view, and evaluating them directly on unseen perspectives without further adaptation. We benchmark simple linear probes, advanced probing strategies, and compare two foundation models (DINOv2 and CLIP) against parameter-efficient fine-tuning (PEFT) and full fine-tuning. Building on these insights, we introduce \textsc{T-Mask} -- a new image-to-video probing method that leverages temporal token masking and emphasizes more dynamic video regions. Benchmarked on the public Drive\&Act dataset, \textsc{T-Mask} improves cross-view top-1 accuracy by $+1.23\%$ over strong probing baselines and $+8.0\%$ over PEFT methods, without adding any parameters. It proves particularly effective for underrepresented secondary activities, boosting recognition by $+5.42\%$ under the trained view and $+1.36\%$ under cross-view settings. This work provides encouraging evidence that adapting foundation models with lightweight probing methods like \textsc{T-Mask} has strong potential in fine-grained driver observation, especially in cross-view and low-data settings. These results highlight the importance of temporal token selection when leveraging foundation models to build robust driver monitoring systems. Code and models will be made available at https://github.com/th-nesh/T-MASK to support ongoing research.
Authors: Shikang Zheng, Liang Feng, Xinyu Wang, Qinming Zhou, Peiliang Cai, Chang Zou, Jiacheng Liu, Yuqi Lin, Junjie Chen, Yue Ma, Linfeng Zhang
Abstract: Diffusion Transformers (DiTs) have demonstrated exceptional performance in high-fidelity image and video generation. To reduce their substantial computational costs, feature caching techniques have been proposed to accelerate inference by reusing hidden representations from previous timesteps. However, current methods often struggle to maintain generation quality at high acceleration ratios, where prediction errors increase sharply due to the inherent instability of long-step forecasting. In this work, we adopt an ordinary differential equation (ODE) perspective on the hidden-feature sequence, modeling layer representations along the trajectory as a feature-ODE. We attribute the degradation of existing caching strategies to their inability to robustly integrate historical features under large skipping intervals. To address this, we propose FoCa (Forecast-then-Calibrate), which treats feature caching as a feature-ODE solving problem. Extensive experiments on image synthesis, video generation, and super-resolution tasks demonstrate the effectiveness of FoCa, especially under aggressive acceleration. Without additional training, FoCa achieves near-lossless speedups of 5.50 times on FLUX, 6.45 times on HunyuanVideo, 3.17 times on Inf-DiT, and maintains high quality with a 4.53 times speedup on DiT.
Authors: Huanpeng Chu, Wei Wu, Guanyu Fen, Yutao Zhang
Abstract: Diffusion models have emerged as a powerful paradigm for generative tasks such as image synthesis and video generation, with Transformer architectures further enhancing performance. However, the high computational cost of diffusion Transformers-stemming from a large number of sampling steps and complex per-step computations-presents significant challenges for real-time deployment. In this paper, we introduce OmniCache, a training-free acceleration method that exploits the global redundancy inherent in the denoising process. Unlike existing methods that determine caching strategies based on inter-step similarities and tend to prioritize reusing later sampling steps, our approach originates from the sampling perspective of DIT models. We systematically analyze the model's sampling trajectories and strategically distribute cache reuse across the entire sampling process. This global perspective enables more effective utilization of cached computations throughout the diffusion trajectory, rather than concentrating reuse within limited segments of the sampling procedure.In addition, during cache reuse, we dynamically estimate the corresponding noise and filter it out to reduce its impact on the sampling direction.Extensive experiments demonstrate that our approach accelerates the sampling process while maintaining competitive generative quality, offering a promising and practical solution for efficient deployment of diffusion-based generative models.
Authors: Kaiyuan Ji, Yijin Guo, Zicheng Zhang, Xiangyang Zhu, Yuan Tian, Ning Liu, Guangtao Zhai
Abstract: With the increasing use of large language models (LLMs) in medical decision-support, it is essential to evaluate not only their final answers but also the reliability of their reasoning. Two key risks are Chain-of-Thought (CoT) faithfulness -- whether reasoning aligns with responses and medical facts -- and sycophancy, where models follow misleading cues over correctness. Existing benchmarks often collapse such vulnerabilities into single accuracy scores. To address this, we introduce MedOmni-45 Degrees, a benchmark and workflow designed to quantify safety-performance trade-offs under manipulative hint conditions. It contains 1,804 reasoning-focused medical questions across six specialties and three task types, including 500 from MedMCQA. Each question is paired with seven manipulative hint types and a no-hint baseline, producing about 27K inputs. We evaluate seven LLMs spanning open- vs. closed-source, general-purpose vs. medical, and base vs. reasoning-enhanced models, totaling over 189K inferences. Three metrics -- Accuracy, CoT-Faithfulness, and Anti-Sycophancy -- are combined into a composite score visualized with a 45 Degrees plot. Results show a consistent safety-performance trade-off, with no model surpassing the diagonal. The open-source QwQ-32B performs closest (43.81 Degrees), balancing safety and accuracy but not leading in both. MedOmni-45 Degrees thus provides a focused benchmark for exposing reasoning vulnerabilities in medical LLMs and guiding safer model development.
Authors: Hohyun Na, Seunghoo Hong, Simon S. Woo
Abstract: The success of diffusion models has enabled effortless, high-quality image modifications that precisely align with users' intentions, thereby raising concerns about their potential misuse by malicious actors. Previous studies have attempted to mitigate such misuse through adversarial attacks. However, these approaches heavily rely on image-level inconsistencies, which pose fundamental limitations in addressing the influence of textual prompts. In this paper, we propose PromptFlare, a novel adversarial protection method designed to protect images from malicious modifications facilitated by diffusion-based inpainting models. Our approach leverages the cross-attention mechanism to exploit the intrinsic properties of prompt embeddings. Specifically, we identify and target shared token of prompts that is invariant and semantically uninformative, injecting adversarial noise to suppress the sampling process. The injected noise acts as a cross-attention decoy, diverting the model's focus away from meaningful prompt-image alignments and thereby neutralizing the effect of prompt. Extensive experiments on the EditBench dataset demonstrate that our method achieves state-of-the-art performance across various metrics while significantly reducing computational overhead and GPU memory usage. These findings highlight PromptFlare as a robust and efficient protection against unauthorized image manipulations. The code is available at https://github.com/NAHOHYUN-SKKU/PromptFlare.
Authors: Sandeep Gupta, Roberto Passerone
Abstract: Visual Foundation Models (VFMs) are becoming ubiquitous in computer vision, powering systems for diverse tasks such as object detection, image classification, segmentation, pose estimation, and motion tracking. VFMs are capitalizing on seminal innovations in deep learning models, such as LeNet-5, AlexNet, ResNet, VGGNet, InceptionNet, DenseNet, YOLO, and ViT, to deliver superior performance across a range of critical computer vision applications. These include security-sensitive domains like biometric verification, autonomous vehicle perception, and medical image analysis, where robustness is essential to fostering trust between technology and the end-users. This article investigates network robustness requirements crucial in computer vision systems to adapt effectively to dynamic environments influenced by factors such as lighting, weather conditions, and sensor characteristics. We examine the prevalent empirical defenses and robust training employed to enhance vision network robustness against real-world challenges such as distributional shifts, noisy and spatially distorted inputs, and adversarial attacks. Subsequently, we provide a comprehensive analysis of the challenges associated with these defense mechanisms, including network properties and components to guide ablation studies and benchmarking metrics to evaluate network robustness.
Authors: Jiahao Chen, Zhiyong Ma, Wenbiao Du, Qingyuan Chuai
Abstract: Multi-modal creative writing (MMCW) aims to produce illustrated articles. Unlike common multi-modal generative (MMG) tasks such as storytelling or caption generation, MMCW is an entirely new and more abstract challenge where textual and visual contexts are not strictly related to each other. Existing methods for related tasks can be forcibly migrated to this track, but they require specific modality inputs or costly training, and often suffer from semantic inconsistencies between modalities. Therefore, the main challenge lies in economically performing MMCW with flexible interactive patterns, where the semantics between the modalities of the output are more aligned. In this work, we propose FlexMUSE with a T2I module to enable optional visual input. FlexMUSE promotes creativity and emphasizes the unification between modalities by proposing the modality semantic alignment gating (msaGate) to restrict the textual input. Besides, an attention-based cross-modality fusion is proposed to augment the input features for semantic enhancement. The modality semantic creative direct preference optimization (mscDPO) within FlexMUSE is designed by extending the rejected samples to facilitate the writing creativity. Moreover, to advance the MMCW, we expose a dataset called ArtMUSE which contains with around 3k calibrated text-image pairs. FlexMUSE achieves promising results, demonstrating its consistency, creativity and coherence.
Authors: Nan wang, Zhiyi Xia, Yiming Li, Shi Tang, Zuxin Fan, Xi Fang, Haoyi Tao, Xiaochen Cai, Guolin Ke, Linfeng Zhang, Yanhui Hong
Abstract: Quantitative microstructural characterization is fundamental to materials science, where electron micrograph (EM) provides indispensable high-resolution insights. However, progress in deep learning-based EM characterization has been hampered by the scarcity of large-scale, diverse, and expert-annotated datasets, due to acquisition costs, privacy concerns, and annotation complexity. To address this issue, we introduce UniEM-3M, the first large-scale and multimodal EM dataset for instance-level understanding. It comprises 5,091 high-resolution EMs, about 3 million instance segmentation labels, and image-level attribute-disentangled textual descriptions, a subset of which will be made publicly available. Furthermore, we are also releasing a text-to-image diffusion model trained on the entire collection to serve as both a powerful data augmentation tool and a proxy for the complete data distribution. To establish a rigorous benchmark, we evaluate various representative instance segmentation methods on the complete UniEM-3M and present UniEM-Net as a strong baseline model. Quantitative experiments demonstrate that this flow-based model outperforms other advanced methods on this challenging benchmark. Our multifaceted release of a partial dataset, a generative model, and a comprehensive benchmark -- available at huggingface -- will significantly accelerate progress in automated materials analysis.
Authors: Yi Xu, Yesheng Zhang, jiajia Liu, Jingdong Chen
Abstract: Multimodal large language models (MLLMs) have emerged as pivotal tools in enhancing human-computer interaction. In this paper we focus on the application of MLLMs in the field of graphical user interface (GUI) elements structuring, where they assist in processing user instructions based on screen contents. Despite the promise of MLLMs, their performance in precisely generating UI element coordinates, a critical aspect of GUI understanding, is hindered by the nature of next-token prediction training. This challenge arises from the semantic void surrounding numerical UI coordinates in language representation spaces, necessitating a substantial and diverse dataset to bolster visual module capabilities. To address these limitations, we introduce an IoU-Augmented Maximum Likelihood (IAML) training paradigm. Specifically, our approach involves a novel pipeline for IoU-based coordinate sampling to augment the training data, which considers the proximity to ground truth coordinates. This data augmentation strategy is then employed to fine-tune MLLMs under the IAML paradigm, which is designed to mitigate the exposure bias problem inherent in traditional maximum likelihood estimation. Through extensive experiments, we demonstrate the superior performance of our IAML training approach over traditional training paradigms.
Authors: Yu Meng, Ligao Deng, Zhihao Xi, Jiansheng Chen, Jingbo Chen, Anzhi Yue, Diyou Liu, Kai Li, Chenhao Wang, Kaiyu Li, Yupeng Deng, Xian Sun
Abstract: With the enhancement of remote sensing image resolution and the rapid advancement of deep learning, land cover mapping is transitioning from pixel-level segmentation to object-based vector modeling. This shift demands more from deep learning models, requiring precise object boundaries and topological consistency. However, existing datasets face three main challenges: limited class annotations, small data scale, and lack of spatial structural information. To overcome these issues, we introduce IRSAMap, the first global remote sensing dataset for large-scale, high-resolution, multi-feature land cover vector mapping. IRSAMap offers four key advantages: 1) a comprehensive vector annotation system with over 1.8 million instances of 10 typical objects (e.g., buildings, roads, rivers), ensuring semantic and spatial accuracy; 2) an intelligent annotation workflow combining manual and AI-based methods to improve efficiency and consistency; 3) global coverage across 79 regions in six continents, totaling over 1,000 km; and 4) multi-task adaptability for tasks like pixel-level classification, building outline extraction, road centerline extraction, and panoramic segmentation. IRSAMap provides a standardized benchmark for the shift from pixel-based to object-based approaches, advancing geographic feature automation and collaborative modeling. It is valuable for global geographic information updates and digital twin construction. The dataset is publicly available at https://github.com/ucas-dlg/IRSAMap
Authors: Khai Duc Minh Tran, Hoa Van Nguyen, Aimuni Binti Muhammad Rawi, Hareeshrao Athinarayanarao, Ba-Ngu Vo
Abstract: This paper tackles the challenging problem of detecting methane plumes, a potent greenhouse gas, using Sentinel-2 imagery. This contributes to the mitigation of rapid climate change. We propose a novel deep learning solution based on U-Net with a ResNet34 encoder, integrating dual spectral enhancement techniques (Varon ratio and Sanchez regression) to optimise input features for heightened sensitivity. A key achievement is the ability to detect small plumes down to 400 m2 (i.e., for a single pixel at 20 m resolution), surpassing traditional methods limited to larger plumes. Experiments show our approach achieves a 78.39% F1-score on the validation set, demonstrating superior performance in sensitivity and precision over existing remote sensing techniques for automated methane monitoring, especially for small plumes.
Authors: Anjith George, Sebastien Marcel
Abstract: The widespread availability of tools for manipulating images and documents has made it increasingly easy to forge digital documents, posing a serious threat to Know Your Customer (KYC) processes and remote onboarding systems. Detecting such forgeries is essential to preserving the integrity and security of these services. In this work, we present EdgeDoc, a novel approach for the detection and localization of document forgeries. Our architecture combines a lightweight convolutional transformer with auxiliary noiseprint features extracted from the images, enhancing its ability to detect subtle manipulations. EdgeDoc achieved third place in the ICCV 2025 DeepID Challenge, demonstrating its competitiveness. Experimental results on the FantasyID dataset show that our method outperforms baseline approaches, highlighting its effectiveness in realworld scenarios. Project page : https://www.idiap. ch/paper/edgedoc/
URLs: https://www.idiap.
Authors: Fengshun Wang, Qiurui Wang, Peilin Zhao
Abstract: Technical Element Score (TES) and Program Component Score (PCS) evaluations in figure skating demand precise assessment of athletic actions and artistic interpretation, respectively. Existing methods face three major challenges. Firstly, video and audio cues are regarded as common features for both TES and PCS predictions in previous works without considering the prior evaluation criterion of figure skating. Secondly, action elements in competitions are separated in time, TES should be derived from each element's score, but existing methods try to give an overall TES prediction without evaluating each action element. Thirdly, lengthy competition videos make it difficult and inefficient to handle long-range contexts. To address these challenges, we propose a two-stream Mamba pyramid network that aligns with actual judging criteria to predict TES and PCS by separating visual-feature based TES evaluation stream from audio-visual-feature based PCS evaluation stream. In the PCS evaluation stream, we introduce a multi-level fusion mechanism to guarantee that video-based features remain unaffected when assessing TES, and enhance PCS estimation by fusing visual and auditory cues across each contextual level of the pyramid. In the TES evaluation stream, the multi-scale Mamba pyramid and TES head we proposed effectively address the challenges of localizing and evaluating action elements with various temporal scales and give score predictions. With Mamba's superior ability to capture long-range dependencies and its linear computational complexity, our method is ideal for handling lengthy figure skating videos. Comprehensive experimentation demonstrates that our framework attains state-of-the-art performance on the FineFS benchmark. Our source code is available at https://github.com/ycwfs/Figure-Skating-Action-Quality-Assessment.
URLs: https://github.com/ycwfs/Figure-Skating-Action-Quality-Assessment.
Authors: Junaid Ahmed Sifat, Abir Chowdhury, Hasnat Md. Imtiaz, Md. Irtiza Hossain, Md. Imran Bin Azad
Abstract: The digitization of handwritten marksheets presents huge challenges due to the different styles of handwriting and complex table structures in such documents like marksheets. This work introduces a hybrid method that integrates OpenCV for table detection and PaddleOCR for recognizing sequential handwritten text. The image processing capabilities of OpenCV efficiently detects rows and columns which enable computationally lightweight and accurate table detection. Additionally, YOLOv8 and Modified YOLOv8 are implemented for handwritten text recognition within the detected table structures alongside PaddleOCR which further enhance the system's versatility. The proposed model achieves high accuracy on our custom dataset which is designed to represent different and diverse handwriting styles and complex table layouts. Experimental results demonstrate that YOLOv8 Modified achieves an accuracy of 92.72 percent, outperforming PaddleOCR 91.37 percent and the YOLOv8 model 88.91 percent. This efficiency reduces the necessity for manual work which makes this a practical and fast solution for digitizing academic as well as administrative documents. This research serves the field of document automation, particularly handwritten document understanding, by providing operational and reliable methods to scale, enhance, and integrate the technologies involved.
Authors: Mohammad Zia Ur Rehman, Devraj Raghuvanshi, Umang Jain, Shubhi Bansal, Nagendra Kumar
Abstract: A major challenge in multimodal learning is the presence of noise within individual modalities. This noise inherently affects the resulting multimodal representations, especially when these representations are obtained through explicit interactions between different modalities. Moreover, the multimodal fusion techniques while aiming to achieve a strong joint representation, can neglect valuable discriminative information within the individual modalities. To this end, we propose a Multimodal-Multitask framework with crOss-modal Relation and hIErarchical iNteractive aTtention (MM-ORIENT) that is effective for multiple tasks. The proposed approach acquires multimodal representations cross-modally without explicit interaction between different modalities, reducing the noise effect at the latent stage. To achieve this, we propose cross-modal relation graphs that reconstruct monomodal features to acquire multimodal representations. The features are reconstructed based on the node neighborhood, where the neighborhood is decided by the features of a different modality. We also propose Hierarchical Interactive Monomadal Attention (HIMA) to focus on pertinent information within a modality. While cross-modal relation graphs help comprehend high-order relationships between two modalities, HIMA helps in multitasking by learning discriminative features of individual modalities before late-fusing them. Finally, extensive experimental evaluation on three datasets demonstrates that the proposed approach effectively comprehends multimodal content for multiple tasks.
Authors: Lucas Maisonnave, Karim Haroun, Tom Pegeot
Abstract: Transformer models rely on Multi-Head Self-Attention (MHSA) mechanisms, where each attention head contributes to the final representation. However, their computational complexity and high memory demands due to MHSA hinders their deployment at the edge. In this work, we analyze and exploit information redundancy in attention maps to accelerate model inference. By quantifying the information captured by each attention head using Shannon entropy, our analysis reveals that attention heads with lower entropy, i.e., exhibiting more deterministic behavior, tend to contribute less information, motivating targeted compression strategies. Relying on these insights, we propose Entropy Attention Maps (EAM), a model that freezes the weights of low-entropy attention maps and quantizes these values to low precision to avoid redundant re-computation. Empirical validation on ImageNet-1k shows that EAM achieves similar or higher accuracy at $\leq$20\% sparsity in attention maps and competitive performance beyond this level for the DeiT and Swin Transformer models.
Authors: Nedyalko Prisadnikov, Danda Pani Paudel, Yuqian Fu, Luc Van Gool
Abstract: This position paper argues that the next generation of vision encoders should be image size agnostic and task driven. The source of our inspiration is biological. Not a structural aspect of biological vision, but a behavioral trait -- efficiency. We focus on a couple of ways in which vision in nature is efficient, but modern vision encoders not. We -- humans and animals -- deal with vast quantities of visual data, and need to be smart where we focus our limited energy -- it depends on the task. It is our belief that vision encoders should be dynamic and the computational complexity should depend on the task at hand rather than the size of the image. We, also, provide concrete first steps towards our vision -- a proof-of-concept solution for image classification. Despite classification being not very representative for what we are trying to achieve, it shows that our approach is feasible and promising.
Authors: Yutaro Hiraoka, Kazuya Okamura, Kota Suto, Kazuhiro Fukui
Abstract: This paper reveals that we can interpret the fundamental function of Randomized Time Warping (RTW) as a type of self-attention mechanism, a core technology of Transformers in motion recognition. The self-attention is a mechanism that enables models to identify and weigh the importance of different parts of an input sequential pattern. On the other hand, RTW is a general extension of Dynamic Time Warping (DTW), a technique commonly used for matching and comparing sequential patterns. In essence, RTW searches for optimal contribution weights for each element of the input sequential patterns to produce discriminative features. Although the two approaches look different, these contribution weights can be interpreted as self-attention weights. In fact, the two weight patterns look similar, producing a high average correlation of 0.80 across the ten smallest canonical angles. However, they work in different ways: RTW attention operates on an entire input sequential pattern, while self-attention focuses on only a local view which is a subset of the input sequential pattern because of the computational costs of the self-attention matrix. This targeting difference leads to an advantage of RTW against Transformer, as demonstrated by the 5\% performance improvement on the Something-Something V2 dataset.
Authors: Yong Zhang, Cunjian Chen, Qiang Gao, Yi Wang, Bin Fang
Abstract: Real-time surface defect detection is critical for maintaining product quality and production efficiency in the steel manufacturing industry. Despite promising accuracy, existing deep learning methods often suffer from high computational complexity and slow inference speeds, which limit their deployment in resource-constrained industrial environments. Recent lightweight approaches adopt multibranch architectures based on depthwise separable convolution (DSConv) to capture multiscale contextual information. However, these methods often suffer from increased computational overhead and lack effective cross-scale feature interaction, limiting their ability to fully leverage multiscale representations. To address these challenges, we propose GMBINet, a lightweight framework that enhances multiscale feature extraction and interaction through novel Group Multiscale Bidirectional Interactive (GMBI) modules. The GMBI adopts a group-wise strategy for multiscale feature extraction, ensuring scale-agnostic computational complexity. It further integrates a Bidirectional Progressive Feature Interactor (BPFI) and a parameter-free Element-Wise Multiplication-Summation (EWMS) operation to enhance cross-scale interaction without introducing additional computational overhead. Experiments on SD-Saliency-900 and NRSD-MN datasets demonstrate that GMBINet delivers competitive accuracy with real-time speeds of 1048 FPS on GPU and 16.53 FPS on CPU at 512 resolution, using only 0.19 M parameters. Additional evaluations on the NEU-CLS defect classification dataset further confirm the strong generalization ability of our method, demonstrating its potential for broader industrial vision applications beyond surface defect detection. The dataset and code are publicly available at: https://github.com/zhangyongcode/GMBINet.
Authors: Edoardo Palladin, Roland Dietze, Praveen Narayanan, Mario Bijelic, Felix Heide
Abstract: Multimodal sensor fusion is an essential capability for autonomous robots, enabling object detection and decision-making in the presence of failing or uncertain inputs. While recent fusion methods excel in normal environmental conditions, these approaches fail in adverse weather, e.g., heavy fog, snow, or obstructions due to soiling. We introduce a novel multi-sensor fusion approach tailored to adverse weather conditions. In addition to fusing RGB and LiDAR sensors, which are employed in recent autonomous driving literature, our sensor fusion stack is also capable of learning from NIR gated camera and radar modalities to tackle low light and inclement weather. We fuse multimodal sensor data through attentive, depth-based blending schemes, with learned refinement on the Bird's Eye View (BEV) plane to combine image and range features effectively. Our detections are predicted by a transformer decoder that weighs modalities based on distance and visibility. We demonstrate that our method improves the reliability of multimodal sensor fusion in autonomous vehicles under challenging weather conditions, bridging the gap between ideal conditions and real-world edge cases. Our approach improves average precision by 17.2 AP compared to the next best method for vulnerable pedestrians in long distances and challenging foggy scenes. Our project page is available at https://light.princeton.edu/samfusion/
Authors: Sara Rojas, Matthieu Armando, Bernard Ghamen, Philippe Weinzaepfel, Vincent Leroy, Gregory Rogez
Abstract: Recovering the 3D geometry of a scene from a sparse set of uncalibrated images is a long-standing problem in computer vision. While recent learning-based approaches such as DUSt3R and MASt3R have demonstrated impressive results by directly predicting dense scene geometry, they are primarily trained on outdoor scenes with static environments and struggle to handle human-centric scenarios. In this work, we introduce HAMSt3R, an extension of MASt3R for joint human and scene 3D reconstruction from sparse, uncalibrated multi-view images. First, we exploit DUNE, a strong image encoder obtained by distilling, among others, the encoders from MASt3R and from a state-of-the-art Human Mesh Recovery (HMR) model, multi-HMR, for a better understanding of scene geometry and human bodies. Our method then incorporates additional network heads to segment people, estimate dense correspondences via DensePose, and predict depth in human-centric environments, enabling a more comprehensive 3D reconstruction. By leveraging the outputs of our different heads, HAMSt3R produces a dense point map enriched with human semantic information in 3D. Unlike existing methods that rely on complex optimization pipelines, our approach is fully feed-forward and efficient, making it suitable for real-world applications. We evaluate our model on EgoHumans and EgoExo4D, two challenging benchmarks con taining diverse human-centric scenarios. Additionally, we validate its generalization to traditional multi-view stereo and multi-view pose regression tasks. Our results demonstrate that our method can reconstruct humans effectively while preserving strong performance in general 3D reconstruction tasks, bridging the gap between human and scene understanding in 3D vision.
Authors: Anilkumar Swamy, Vincent Leroy, Philippe Weinzaepfel, Jean-S\'ebastien Franco, Gr\'egory Rogez
Abstract: Hand-object 3D reconstruction has become increasingly important for applications in human-robot interaction and immersive AR/VR experiences. A common approach for object-agnostic hand-object reconstruction from RGB sequences involves a two-stage pipeline: hand-object 3D tracking followed by multi-view 3D reconstruction. However, existing methods rely on keypoint detection techniques, such as Structure from Motion (SfM) and hand-keypoint optimization, which struggle with diverse object geometries, weak textures, and mutual hand-object occlusions, limiting scalability and generalization. As a key enabler to generic and seamless, non-intrusive applicability, we propose in this work a robust, keypoint detector-free approach to estimating hand-object 3D transformations from monocular motion video/images. We further integrate this with a multi-view reconstruction pipeline to accurately recover hand-object 3D shape. Our method, named HOSt3R, is unconstrained, does not rely on pre-scanned object templates or camera intrinsics, and reaches state-of-the-art performance for the tasks of object-agnostic hand-object 3D transformation and shape estimation on the SHOWMe benchmark. We also experiment on sequences from the HO3D dataset, demonstrating generalization to unseen object categories.
Authors: Huimin Zeng, Yue Bai, Yun Fu
Abstract: Existing 3D Gaussian Splatting (3DGS) super-resolution methods typically perform high-resolution (HR) rendering of fixed scale factors, making them impractical for resource-limited scenarios. Directly rendering arbitrary-scale HR views with vanilla 3DGS introduces aliasing artifacts due to the lack of scale-aware rendering ability, while adding a post-processing upsampler for 3DGS complicates the framework and reduces rendering efficiency. To tackle these issues, we build an integrated framework that incorporates scale-aware rendering, generative prior-guided optimization, and progressive super-resolving to enable 3D Gaussian super-resolution of arbitrary scale factors with a single 3D model. Notably, our approach supports both integer and non-integer scale rendering to provide more flexibility. Extensive experiments demonstrate the effectiveness of our model in rendering high-quality arbitrary-scale HR views (6.59 dB PSNR gain over 3DGS) with a single model. It preserves structural consistency with LR views and across different scales, while maintaining real-time rendering speed (85 FPS at 1080p).
Authors: Chun-Peng Chang, Chen-Yu Wang, Julian Schmidt, Holger Caesar, Alain Pagani
Abstract: Recent advancements in video generation have substantially improved visual quality and temporal coherence, making these models increasingly appealing for applications such as autonomous driving, particularly in the context of driving simulation and so-called "world models". In this work, we investigate the effects of existing fine-tuning video generation approaches on structured driving datasets and uncover a potential trade-off: although visual fidelity improves, spatial accuracy in modeling dynamic elements may degrade. We attribute this degradation to a shift in the alignment between visual quality and dynamic understanding objectives. In datasets with diverse scene structures within temporal space, where objects or perspective shift in varied ways, these objectives tend to highly correlated. However, the very regular and repetitive nature of driving scenes allows visual quality to improve by modeling dominant scene motion patterns, without necessarily preserving fine-grained dynamic behavior. As a result, fine-tuning encourages the model to prioritize surface-level realism over dynamic accuracy. To further examine this phenomenon, we show that simple continual learning strategies, such as replay from diverse domains, can offer a balanced alternative by preserving spatial accuracy while maintaining strong visual quality.
Authors: Andrei-Stefan Bulzan, Cosmin Cernazanu-Glavan
Abstract: For decades, Computer Vision has aimed at enabling machines to perceive the external world. Initial limitations led to the development of highly specialized niches. As success in each task accrued and research progressed, increasingly complex perception tasks emerged. This survey charts the convergence of these tasks and, in doing so, introduces Open World Detection (OWD), an umbrella term we propose to unify class-agnostic and generally applicable detection models in the vision domain. We start from the history of foundational vision subdomains and cover key concepts, methodologies and datasets making up today's state-of-the-art landscape. This traverses topics starting from early saliency detection, foreground/background separation, out of distribution detection and leading up to open world object detection, zero-shot detection and Vision Large Language Models (VLLMs). We explore the overlap between these subdomains, their increasing convergence, and their potential to unify into a singular domain in the future, perception.
Authors: Yosef Dayani, Omer Benishu, Sagie Benaim
Abstract: Text-to-3D generation approaches have advanced significantly by leveraging pretrained 2D diffusion priors, producing high-quality and 3D-consistent outputs. However, they often fail to produce out-of-domain (OOD) or rare concepts, yielding inconsistent or inaccurate results. To this end, we propose MV-RAG, a novel text-to-3D pipeline that first retrieves relevant 2D images from a large in-the-wild 2D database and then conditions a multiview diffusion model on these images to synthesize consistent and accurate multiview outputs. Training such a retrieval-conditioned model is achieved via a novel hybrid strategy bridging structured multiview data and diverse 2D image collections. This involves training on multiview data using augmented conditioning views that simulate retrieval variance for view-specific reconstruction, alongside training on sets of retrieved real-world 2D images using a distinctive held-out view prediction objective: the model predicts the held-out view from the other views to infer 3D consistency from 2D data. To facilitate a rigorous OOD evaluation, we introduce a new collection of challenging OOD prompts. Experiments against state-of-the-art text-to-3D, image-to-3D, and personalization baselines show that our approach significantly improves 3D consistency, photorealism, and text adherence for OOD/rare concepts, while maintaining competitive performance on standard benchmarks.
Authors: Ortensia Forni, Alexandre Darmon, Michael Benzaquen
Abstract: While color harmony has long been studied in art and design, a clear consensus remains elusive, as most models are grounded in qualitative insights or limited datasets. In this work, we present a quantitative, data-driven study of color pairing preferences using controlled hue-based palettes in the HSL color space. Participants evaluated combinations of thirteen distinct hues, enabling us to construct a preference matrix and define a combinability index for each color. Our results reveal that preferences are highly hue dependent, challenging the assumption of universal harmony rules proposed in the literature. Yet, when averaged over hues, statistically meaningful patterns of aesthetic preference emerge, with certain hue separations perceived as more harmonious. Strikingly, these patterns align with hue distributions found in natural landscapes, pointing to a statistical correspondence between human color preferences and the structure of color in nature. Together, these findings offer a quantitative framework for studying color harmony and its potential perceptual and ecological underpinnings.
Authors: Xiaoxu Zhu
Abstract: Finite Scalar Quantization (FSQ) has emerged as a promising alternative to Vector Quantization (VQ) in neural compression, offering simplified training and improved stability. However, naive application of FSQ in residual quantization frameworks suffers from the \textbf{residual magnitude decay problem}, where subsequent FSQ layers receive progressively weaker signals, severely limiting their effectiveness. We propose \textbf{Robust Residual Finite Scalar Quantization (RFSQ)}, a general framework that addresses this fundamental limitation through two novel conditioning strategies: learnable scaling factors and invertible layer normalization. Our approach maintains the simplicity of FSQ while enabling effective multi-stage residual quantization. Comprehensive experiments on ImageNet demonstrate that RFSQ variants significantly outperform strong baselines including VQ-EMA, FSQ, and LFQ, achieving up to 45\% improvement in perceptual loss and 28.7\% reduction in L1 reconstruction error. The proposed LayerNorm strategy shows the most consistent improvements across different configurations, establishing RFSQ as a superior quantization method for neural compression.
Authors: Zhaodong Jiang, Ashish Sinha, Tongtong Cao, Yuan Ren, Bingbing Liu, Binbin Xu
Abstract: Estimating the 6D pose of novel objects is a fundamental yet challenging problem in robotics, often relying on access to object CAD models. However, acquiring such models can be costly and impractical. Recent approaches aim to bypass this requirement by leveraging strong priors from foundation models to reconstruct objects from single or multi-view images, but typically require additional training or produce hallucinated geometry. To this end, we propose UnPose, a novel framework for zero-shot, model-free 6D object pose estimation and reconstruction that exploits 3D priors and uncertainty estimates from a pre-trained diffusion model. Specifically, starting from a single-view RGB-D frame, UnPose uses a multi-view diffusion model to estimate an initial 3D model using 3D Gaussian Splatting (3DGS) representation, along with pixel-wise epistemic uncertainty estimates. As additional observations become available, we incrementally refine the 3DGS model by fusing new views guided by the diffusion model's uncertainty, thereby continuously improving the pose estimation accuracy and 3D reconstruction quality. To ensure global consistency, the diffusion prior-generated views and subsequent observations are further integrated in a pose graph and jointly optimized into a coherent 3DGS field. Extensive experiments demonstrate that UnPose significantly outperforms existing approaches in both 6D pose estimation accuracy and 3D reconstruction quality. We further showcase its practical applicability in real-world robotic manipulation tasks.
Authors: Surajit Das, Pavel Zun
Abstract: Brightfield microscopy imaging of unstained live cells remains a persistent challenge due to low contrast, temporal changes in specimen phenotypes, irregular illumination, and the absence of training labels. While deep learning (DL) methods (e.g., Cellpose 3.0) achieve state-of-the-art (SOTA) performance, they require extensive labeled data and heavy computational resources, and they often fail under uneven illumination. We present the first unsupervised segmentation framework combining spatial standard deviation from local mean (SSDLM), fuzzy logic, adjusted variograms, Moran's I, and cumulative squared shift of nodal intensity (CSSNI) to address these limitations. Unlike deep learning models, our approach requires no annotations or retraining and operates through a user-friendly GUI tailored for non-programming users. The robustness and generality were validated on three datasets, including cross-domain data. We benchmark our method against 2023--2024 SOTA models, including Cellpose 3.0 and StarDist, using a dataset of unstained myoblast images. Our method achieves a significant improvement in segmentation performance, with an IoU increase of up to 48\% and statistically validated superiority ($p < 0.01$, Wilcoxon signed-rank test). Expert evaluation from two biologists further supports the segmentation quality (Cohen's $\kappa > 0.75$). The proposed algorithm is lightweight, interpretable, and computationally efficient, offering a practical and effective alternative for cell segmentation in label-free microscopy. The code, the dataset, and the results are available for reproducibility*.
Authors: Hung-Jui Huang, Mohammad Amin Mirzaee, Michael Kaess, Wenzhen Yuan
Abstract: Accurately perceiving an object's pose and shape is essential for precise grasping and manipulation. Compared to common vision-based methods, tactile sensing offers advantages in precision and immunity to occlusion when tracking and reconstructing objects in contact. This makes it particularly valuable for in-hand and other high-precision manipulation tasks. In this work, we present GelSLAM, a real-time 3D SLAM system that relies solely on tactile sensing to estimate object pose over long periods and reconstruct object shapes with high fidelity. Unlike traditional point cloud-based approaches, GelSLAM uses tactile-derived surface normals and curvatures for robust tracking and loop closure. It can track object motion in real time with low error and minimal drift, and reconstruct shapes with submillimeter accuracy, even for low-texture objects such as wooden tools. GelSLAM extends tactile sensing beyond local contact to enable global, long-horizon spatial perception, and we believe it will serve as a foundation for many precise manipulation tasks involving interaction with objects in hand. The video demo is available on our website: https://joehjhuang.github.io/gelslam.
Authors: Muhaisin Tiyumba Nantogmah, Abdul-Barik Alhassan, Salamudeen Alhassan
Abstract: A precise assessment of the risk of breast lesions can greatly lower it and assist physicians in choosing the best course of action. To categorise breast lesions, the majority of current computer-aided systems only use characteristics from mammograms. Although this method is practical, it does not completely utilise clinical reports' valuable information to attain the best results. When compared to utilising mammography alone, will clinical features greatly enhance the categorisation of breast lesions? How may clinical features and mammograms be combined most effectively? In what ways may explainable AI approaches improve the interpretability and reliability of models used to diagnose breast cancer? To answer these basic problems, a comprehensive investigation is desperately needed. In order to integrate mammography and categorical clinical characteristics, this study examines a number of multimodal deep networks grounded on feature concatenation, co-attention, and cross-attention. The model achieved an AUC-ROC of 0.98, accuracy of 0.96, F1-score of 0.94, precision of 0.92, and recall of 0.95 when tested on publicly accessible datasets (TCGA and CBIS-DDSM).
Authors: Juampablo E. Heras Rivera, Hitender Oswal, Tianyi Ren, Yutong Pan, William Henry, Caitlin M. Neher, Mehmet Kurt
Abstract: Stroke is among the top three causes of death worldwide, and accurate identification of ischemic stroke lesion boundaries from imaging is critical for diagnosis and treatment. The main imaging modalities used include magnetic resonance imaging (MRI), particularly diffusion weighted imaging (DWI), and computed tomography (CT)-based techniques such as non-contrast CT (NCCT), contrast-enhanced CT angiography (CTA), and CT perfusion (CTP). DWI is the gold standard for the identification of lesions but has limited applicability in low-resource settings due to prohibitive costs. CT-based imaging is currently the most practical imaging method in low-resource settings due to low costs and simplified logistics, but lacks the high specificity of MRI-based methods in monitoring ischemic insults. Supervised deep learning methods are the leading solution for automated ischemic stroke lesion segmentation and provide an opportunity to improve diagnostic quality in low-resource settings by incorporating insights from DWI when segmenting from CT. Here, we develop a series of models which use CT images taken upon arrival as inputs to predict follow-up lesion volumes annotated from DWI taken 2-9 days later. Furthermore, we implement clinically motivated preprocessing steps and show that the proposed pipeline results in a 38% improvement in Dice score over 10 folds compared to a nnU-Net model trained with the baseline preprocessing. Finally, we demonstrate that through additional preprocessing of CTA maps to extract vessel segmentations, we further improve our best model by 21% over 5 folds.
Authors: Prateek Poudel, Prashant Aryal, Kirtan Kunwar, Navin Nepal, Dinesh Bania Kshatri
Abstract: We investigate the use of wavelet-space feature decomposition in neural super-resolution for rendering pipelines. Building on the DFASR framework, we introduce a wavelet-domain representation that separates low- and high-frequency details before reconstruction, enabling the network to better preserve fine textures while maintaining structural consistency. Unlike RGB-space regression, our approach leverages the stationary wavelet transform (SWT) to avoid spatial down-sampling, ensuring alignment across subbands and preserving shift invariance. The model predicts wavelet coefficients conditioned on spatial G-buffers and temporally warped history frames, which are then recombined through inverse wavelet synthesis. We conduct a comprehensive ablation study across wavelet families, transform types, and architectural variants, showing that incorporating SWT improves PSNR by up to 1.5 dB and reduces LPIPS by 17% on average, at a computational overhead of roughly +24 ms compared to out DFASR baseline. While absolute runtimes on our RTX 3050 mobile GPU are higher ( 141ms) than the original DFASR report on RTX 4090( 11ms), the relative overhead remains modest, suggesting that on higher-end GPUs our method would also remain real-time capable. Taken together, our results suggest that wavelet-domain representations are a principled and effective way to enhance perceptual quality in neural upscaling for graphics applications.
Authors: Md Tariquzzaman, Md Farhan Ishmam, Saiyma Sittul Muna, Md Kamrul Hasan, Hasan Mahmud
Abstract: Sign Language (SL) enables two-way communication for the deaf and hard-of-hearing community, yet many sign languages remain under-resourced in the AI space. Sign Language Instruction Generation (SLIG) produces step-by-step textual instructions that enable non-SL users to imitate and learn SL gestures, promoting two-way interaction. We introduce BdSLIG, the first Bengali SLIG dataset, used to evaluate Vision Language Models (VLMs) (i) on under-resourced SLIG tasks, and (ii) on long-tail visual concepts, as Bengali SL is unlikely to appear in the VLM pre-training data. To enhance zero-shot performance, we introduce Sign Parameter-Infused (SPI) prompting, which integrates standard SL parameters, like hand shape, motion, and orientation, directly into the textual prompts. Subsuming standard sign parameters into the prompt makes the instructions more structured and reproducible than free-form natural text from vanilla prompting. We envision that our work would promote inclusivity and advancement in SL learning systems for the under-resourced communities.
Authors: Wontae Kim, Keuntek Lee, Nam Ik Cho
Abstract: The image enhancement methods based on 3D lookup tables (3D LUTs) efficiently reduce both model size and runtime by interpolating pre-calculated values at the vertices. However, the 3D LUT methods have a limitation due to their lack of spatial information, as they convert color values on a point-by-point basis. Although spatial-aware 3D LUT methods address this limitation, they introduce additional modules that require a substantial number of parameters, leading to increased runtime as image resolution increases. To address this issue, we propose a method for generating image-adaptive LUTs by focusing on the redundant parts of the tables. Our efficient framework decomposes a 3D LUT into a linear sum of low-dimensional LUTs and employs singular value decomposition (SVD). Furthermore, we enhance the modules for spatial feature fusion to be more cache-efficient. Extensive experimental results demonstrate that our model effectively decreases both the number of parameters and runtime while maintaining spatial awareness and performance.
Authors: Weiting Tan, Jiachen Lian, Hirofumi Inaguma, Paden Tomasello, Philipp Koehn, Xutai Ma
Abstract: We present an Audio-Visual Language Model (AVLM) for expressive speech generation by integrating full-face visual cues into a pre-trained expressive speech model. We explore multiple visual encoders and multimodal fusion strategies during pre-training to identify the most effective integration approach. Subsequent fine-tuning on emotion recognition and expressive dialogue tasks yields substantial gains over speech-only baselines (e.g., +5 F1 in emotion recognition). AVLM highlights the value of expressive visual information in guiding speech generation and offers a foundation for end-to-end multimodal conversational systems.
Authors: Yijie Zhang, Cagatay Isil, Xilin Yang, Yuzhu Li, Anna Elia, Karin Atlan, William Dean Wallace, Nir Pillar, Aydogan Ozcan
Abstract: Immunohistochemistry (IHC) has transformed clinical pathology by enabling the visualization of specific proteins within tissue sections. However, traditional IHC requires one tissue section per stain, exhibits section-to-section variability, and incurs high costs and laborious staining procedures. While multiplexed IHC (mIHC) techniques enable simultaneous staining with multiple antibodies on a single slide, they are more tedious to perform and are currently unavailable in routine pathology laboratories. Here, we present a deep learning-based virtual multiplexed immunostaining framework to simultaneously generate ERG and PanCK, in addition to H&E virtual staining, enabling accurate localization and interpretation of vascular invasion in thyroid cancers. This virtual mIHC technique is based on the autofluorescence microscopy images of label-free tissue sections, and its output images closely match the histochemical staining counterparts (ERG, PanCK and H&E) of the same tissue sections. Blind evaluation by board-certified pathologists demonstrated that virtual mIHC staining achieved high concordance with the histochemical staining results, accurately highlighting epithelial cells and endothelial cells. Virtual mIHC conducted on the same tissue section also allowed the identification and localization of small vessel invasion. This multiplexed virtual IHC approach can significantly improve diagnostic accuracy and efficiency in the histopathological evaluation of vascular invasion, potentially eliminating the need for traditional staining protocols and mitigating issues related to tissue loss and heterogeneity.
Authors: Philipp D. L\"osel, Aleese Barron, Yulai Zhang, Matthias Fabian, Benjamin Young, Nicolas Francois, Andrew M. Kingston
Abstract: Non-destructive 3D imaging of large multi-particulate samples is essential for quantifying particle-level properties, such as size, shape, and spatial distribution, across applications in mining, materials science, and geology. However, accurate instance segmentation of particles in tomographic data remains challenging due to high morphological variability and frequent particle contact, which limit the effectiveness of classical methods like watershed algorithms. While supervised deep learning approaches offer improved performance, they rely on extensive annotated datasets that are labor-intensive, error-prone, and difficult to scale. In this work, we propose self-validated learning, a novel self-training framework for particle instance segmentation that eliminates the need for manual annotations. Our method leverages implicit boundary detection and iteratively refines the training set by identifying particles that can be consistently matched across reshuffled scans of the same sample. This self-validation mechanism mitigates the impact of noisy pseudo-labels, enabling robust learning from unlabeled data. After just three iterations, our approach accurately segments over 97% of the total particle volume and identifies more than 54,000 individual particles in tomographic scans of quartz fragments. Importantly, the framework also enables fully autonomous model evaluation without the need for ground truth annotations, as confirmed through comparisons with state-of-the-art instance segmentation techniques. The method is integrated into the Biomedisa image analysis platform (https://github.com/biomedisa/biomedisa/).
Authors: H\'el\`ene Corbaz, Anh Nguyen, Victor Schulze-Zachau, Paul Friedrich, Alicia Durrer, Florentin Bieder, Philippe C. Cattin, Marios N Psychogios
Abstract: Patients undergoing a mechanical thrombectomy procedure usually have a multi-detector CT (MDCT) scan before and after the intervention. The image quality of the flat panel detector CT (FDCT) present in the intervention room is generally much lower than that of a MDCT due to significant artifacts. However, using only FDCT images could improve patient management as the patient would not need to be moved to the MDCT room. Several studies have evaluated the potential use of FDCT imaging alone and the time that could be saved by acquiring the images before and/or after the intervention only with the FDCT. This study proposes using a denoising diffusion probabilistic model (DDPM) to improve the image quality of FDCT scans, making them comparable to MDCT scans. Clinicans evaluated FDCT, MDCT, and our model's predictions for diagnostic purposes using a questionnaire. The DDPM eliminated most artifacts and improved anatomical visibility without reducing bleeding detection, provided that the input FDCT image quality is not too low. Our code can be found on github.
Authors: Odin Hoff Gardaa, Nello Blaser
Abstract: Contours or closed planar curves are common in many domains. For example, they appear as object boundaries in computer vision, isolines in meteorology, and the orbits of rotating machinery. In many cases when learning from contour data, planar rotations of the input will result in correspondingly rotated outputs. It is therefore desirable that deep learning models be rotationally equivariant. In addition, contours are typically represented as an ordered sequence of edge points, where the choice of starting point is arbitrary. It is therefore also desirable for deep learning methods to be equivariant under cyclic shifts. We present RotaTouille, a deep learning framework for learning from contour data that achieves both rotation and cyclic shift equivariance through complex-valued circular convolution. We further introduce and characterize equivariant non-linearities, coarsening layers, and global pooling layers to obtain invariant representations for downstream tasks. Finally, we demonstrate the effectiveness of RotaTouille through experiments in shape classification, reconstruction, and contour regression.
Authors: Badhan Mazumder, Aline Kotoski, Vince D. Calhoun, Dong Hye Ye
Abstract: Understanding how prenatal exposure to psychoactive substances such as cannabis shapes adolescent brain organization remains a critical challenge, complicated by the complexity of multimodal neuroimaging data and the limitations of conventional analytic methods. Existing approaches often fail to fully capture the complementary features embedded within structural and functional connectomes, constraining both biological insight and predictive performance. To address this, we introduced NeuroKoop, a novel graph neural network-based framework that integrates structural and functional brain networks utilizing neural Koopman operator-driven latent space fusion. By leveraging Koopman theory, NeuroKoop unifies node embeddings derived from source-based morphometry (SBM) and functional network connectivity (FNC) based brain graphs, resulting in enhanced representation learning and more robust classification of prenatal drug exposure (PDE) status. Applied to a large adolescent cohort from the ABCD dataset, NeuroKoop outperformed relevant baselines and revealed salient structural-functional connections, advancing our understanding of the neurodevelopmental impact of PDE.
Authors: Hafeez Ur Rehman, Sumaiya Fazal, Moutaz Alazab, Ali Baydoun
Abstract: Glioblastomas, constituting over 50% of malignant brain tumors, are highly aggressive brain tumors that pose substantial treatment challenges due to their rapid progression and resistance to standard therapies. The methylation status of the O-6-Methylguanine-DNA Methyltransferase (MGMT) gene is a critical biomarker for predicting patient response to treatment, particularly with the alkylating agent temozolomide. However, accurately predicting MGMT methylation status using non-invasive imaging techniques remains challenging due to the complex and heterogeneous nature of glioblastomas, that includes, uneven contrast, variability within lesions, and irregular enhancement patterns. This study introduces the Convolutional Autoencoders for MGMT Methylation Status Prediction (CAMP) framework, which is based on adaptive sparse penalties to enhance predictive accuracy. The CAMP framework operates in two phases: first, generating synthetic MRI slices through a tailored autoencoder that effectively captures and preserves intricate tissue and tumor structures across different MRI modalities; second, predicting MGMT methylation status using a convolutional neural network enhanced by adaptive sparse penalties. The adaptive sparse penalty dynamically adjusts to variations in the data, such as contrast differences and tumor locations in MR images. Our method excels in MRI image synthesis, preserving brain tissue, fat, and individual tumor structures across all MRI modalities. Validated on benchmark datasets, CAMP achieved an accuracy of 0.97, specificity of 0.98, and sensitivity of 0.97, significantly outperforming existing methods. These results demonstrate the potential of the CAMP framework to improve the interpretation of MRI data and contribute to more personalized treatment strategies for glioblastoma patients.
Authors: Aniello Panariello, Emanuele Frascaroli, Pietro Buzzega, Lorenzo Bonicelli, Angelo Porrello, Simone Calderara
Abstract: The advent of pre-trained Vision-Language Models (VLMs) has significantly transformed Continual Learning (CL), mainly due to their zero-shot classification abilities. Such proficiency makes VLMs well-suited for real-world applications, enabling robust performance on novel unseen classes without requiring adaptation. However, fine-tuning remains essential when downstream tasks deviate significantly from the pre-training domain. Prior CL approaches primarily focus on preserving the zero-shot capabilities of VLMs during incremental fine-tuning on a downstream task. We take a step further by devising an approach that transforms preservation into enhancement of the zero-shot capabilities of VLMs. Our approach, named MoDular Embedding Recomposition (MoDER), introduces a modular framework that trains multiple textual experts, each specialized in a single seen class, and stores them in a foundational hub. At inference time, for each unseen class, we query the hub and compose the retrieved experts to synthesize a refined prototype that improves classification. We show the effectiveness of our method across two popular zero-shot incremental protocols, Class-IL and MTIL, comprising a total of 14 datasets. The codebase is available at https://github.com/aimagelab/mammoth.
Authors: Yupei Zhang, Xiaofei Wang, Anran Liu, Lequan Yu, Chao Li
Abstract: Histopathology remains the gold standard for cancer diagnosis and prognosis. With the advent of transcriptome profiling, multi-modal learning combining transcriptomics with histology offers more comprehensive information. However, existing multi-modal approaches are challenged by intrinsic multi-modal heterogeneity, insufficient multi-scale integration, and reliance on paired data, restricting clinical applicability. To address these challenges, we propose a disentangled multi-modal framework with four contributions: 1) To mitigate multi-modal heterogeneity, we decompose WSIs and transcriptomes into tumor and microenvironment subspaces using a disentangled multi-modal fusion module, and introduce a confidence-guided gradient coordination strategy to balance subspace optimization. 2) To enhance multi-scale integration, we propose an inter-magnification gene-expression consistency strategy that aligns transcriptomic signals across WSI magnifications. 3) To reduce dependency on paired data, we propose a subspace knowledge distillation strategy enabling transcriptome-agnostic inference through a WSI-only student model. 4) To improve inference efficiency, we propose an informative token aggregation module that suppresses WSI redundancy while preserving subspace semantics. Extensive experiments on cancer diagnosis, prognosis, and survival prediction demonstrate our superiority over state-of-the-art methods across multiple settings. Code is available at https://github.com/helenypzhang/Disentangled-Multimodal-Learning.
URLs: https://github.com/helenypzhang/Disentangled-Multimodal-Learning.
Authors: Tim Langer, Matthias Widra, Volkhard Beyer
Abstract: In the context of industry 4.0, long-serving industrial machines can be retrofitted with process monitoring capabilities for future use in a smart factory. One possible approach is the deployment of wireless monitoring systems, which can benefit substantially from the TinyML paradigm. This work presents a complete TinyML flow from dataset generation, to machine learning model development, up to implementation and evaluation of a full preprocessing and classification pipeline on a microcontroller. After a short review on TinyML in industrial process monitoring, the creation of the novel MillingVibes dataset is described. The feasibility of a TinyML system for structure-integrated process quality monitoring could be shown by the development of an 8-bit-quantized convolutional neural network (CNN) model with 12.59kiB parameter storage. A test accuracy of 100.0% could be reached at 15.4ms inference time and 1.462mJ per quantized CNN inference on an ARM Cortex M4F microcontroller, serving as a reference for future TinyML process monitoring solutions.
Authors: Tainyi Zhang, Zheng-Peng Duan, Peng-Tao Jiang, Bo Li, Ming-Ming Cheng, Chun-Le Guo, Chongyi Li
Abstract: Diffusion-based real-world image super-resolution (Real-ISR) methods have demonstrated impressive performance. To achieve efficient Real-ISR, many works employ Variational Score Distillation (VSD) to distill pre-trained stable-diffusion (SD) model for one-step SR with a fixed timestep. However, due to the different noise injection timesteps, the SD will perform different generative priors. Therefore, a fixed timestep is difficult for these methods to fully leverage the generative priors in SD, leading to suboptimal performance. To address this, we propose a Time-Aware one-step Diffusion Network for Real-ISR (TADSR). We first introduce a Time-Aware VAE Encoder, which projects the same image into different latent features based on timesteps. Through joint dynamic variation of timesteps and latent features, the student model can better align with the input pattern distribution of the pre-trained SD, thereby enabling more effective utilization of SD's generative capabilities. To better activate the generative prior of SD at different timesteps, we propose a Time-Aware VSD loss that bridges the timesteps of the student model and those of the teacher model, thereby producing more consistent generative prior guidance conditioned on timesteps. Additionally, though utilizing the generative prior in SD at different timesteps, our method can naturally achieve controllable trade-offs between fidelity and realism by changing the timestep condition. Experimental results demonstrate that our method achieves both state-of-the-art performance and controllable SR results with only a single step.
Authors: Guangyu Sun, Jingtao Li, Weiming Zhuang, Chen Chen, Chen Chen, Lingjuan Lyu
Abstract: Foundation models (FMs) exhibit remarkable generalization but require adaptation to downstream tasks, particularly in privacy-sensitive applications. Due to data privacy regulations, cloud-based FMs cannot directly access private edge data, limiting their adaptation. Federated learning (FL) provides a privacy-aware alternative, but existing FL approaches overlook the constraints imposed by edge devices -- namely, limited computational resources and the scarcity of labeled data. To address these challenges, we introduce Practical Semi-Supervised Federated Learning (PSSFL), where edge devices hold only unlabeled, low-resolution data, while the server has limited labeled, high-resolution data. In this setting, we propose the Federated Mixture of Experts (FedMox), a novel framework that enhances FM adaptation in FL. FedMox tackles computational and resolution mismatch challenges via a sparse Mixture-of-Experts architecture, employing a spatial router to align features across resolutions and a Soft-Mixture strategy to stabilize semi-supervised learning. We take object detection as a case study, and experiments on real-world autonomous driving datasets demonstrate that FedMox effectively adapts FMs under PSSFL, significantly improving performance with constrained memory costs on edge devices. Our work paves the way for scalable and privacy-preserving FM adaptation in federated scenarios.
Authors: Yuhui Tao, Zhongwei Zhao, Zilong Wang, Xufang Luo, Feng Chen, Kang Wang, Chuanfu Wu, Xue Zhang, Shaoting Zhang, Jiaxi Yao, Xingwei Jin, Xinyang Jiang, Yifan Yang, Dongsheng Li, Lili Qiu, Zhiqiang Shao, Jianming Guo, Nengwang Yu, Shuo Wang, Ying Xiong
Abstract: The non-invasive assessment of increasingly incidentally discovered renal masses is a critical challenge in urologic oncology, where diagnostic uncertainty frequently leads to the overtreatment of benign or indolent tumors. In this study, we developed and validated RenalCLIP using a dataset of 27,866 CT scans from 8,809 patients across nine Chinese medical centers and the public TCIA cohort, a visual-language foundation model for characterization, diagnosis and prognosis of renal mass. The model was developed via a two-stage pre-training strategy that first enhances the image and text encoders with domain-specific knowledge before aligning them through a contrastive learning objective, to create robust representations for superior generalization and diagnostic precision. RenalCLIP achieved better performance and superior generalizability across 10 core tasks spanning the full clinical workflow of kidney cancer, including anatomical assessment, diagnostic classification, and survival prediction, compared with other state-of-the-art general-purpose CT foundation models. Especially, for complicated task like recurrence-free survival prediction in the TCIA cohort, RenalCLIP achieved a C-index of 0.726, representing a substantial improvement of approximately 20% over the leading baselines. Furthermore, RenalCLIP's pre-training imparted remarkable data efficiency; in the diagnostic classification task, it only needs 20% training data to achieve the peak performance of all baseline models even after they were fully fine-tuned on 100% of the data. Additionally, it achieved superior performance in report generation, image-text retrieval and zero-shot diagnosis tasks. Our findings establish that RenalCLIP provides a robust tool with the potential to enhance diagnostic accuracy, refine prognostic stratification, and personalize the management of patients with kidney cancer.
Authors: Yuedong Chen, Haofei Xu, Qianyi Wu, Chuanxia Zheng, Tat-Jen Cham, Jianfei Cai
Abstract: We present a new generalizable NeRF method that is able to directly generalize to new unseen scenarios and perform novel view synthesis with as few as two source views. The key to our approach lies in the explicitly modeled correspondence matching information, so as to provide the geometry prior to the prediction of NeRF color and density for volume rendering. The explicit correspondence matching is quantified with the cosine similarity between image features sampled at the 2D projections of a 3D point on different views, which is able to provide reliable cues about the surface geometry. Unlike previous methods where image features are extracted independently for each view, we consider modeling the cross-view interactions via Transformer cross-attention, which greatly improves the feature matching quality. Our method achieves state-of-the-art results on different evaluation settings, with the experiments showing a strong correlation between our learned cosine feature similarity and volume density, demonstrating the effectiveness and superiority of our proposed method. The code and model are on our project page: https://donydchen.github.io/matchnerf
Authors: Gousia Habib, Tausifa Jan Saleem, Ishfaq Ahmad Malik, Brejesh Lall
Abstract: With the rapid development of computer vision, Vision Transformers (ViTs) offer the tantalising prospect of unified information processing across visual and textual domains due to the lack of inherent inductive biases in ViTs. ViTs require enormous datasets for training. We introduce an innovative ensemble-based distillation approach that distils inductive bias from complementary lightweight teacher models to make their applications practical. Prior systems relied solely on convolution-based teaching. However, this method incorporates an ensemble of light teachers with different architectural tendencies, such as convolution and involution, to jointly instruct the student transformer. Because of these unique inductive biases, instructors can accumulate a wide range of knowledge, even from readily identifiable stored datasets, which leads to enhanced student performance. Our proposed framework LIB-KD also involves precomputing and keeping logits in advance, essentially the unnormalized predictions of the model. This optimisation can accelerate the distillation process by eliminating the need for repeated forward passes during knowledge distillation, significantly reducing the computational burden and enhancing efficiency.
Authors: Yingqi Lin, Xiaogang Xu, Jiafei Wu, Yan Han, Zhe Liu
Abstract: Low-Light Enhancement (LLE) is aimed at improving the quality of photos/videos captured under low-light conditions. It is worth noting that most existing LLE methods do not take advantage of geometric modeling. We believe that incorporating geometric information can enhance LLE performance, as it provides insights into the physical structure of the scene that influences illumination conditions. To address this, we propose a Geometry-Guided Low-Light Enhancement Refine Framework (GG-LLERF) designed to assist low-light enhancement models in learning improved features for LLE by integrating geometric priors into the feature representation space. In this paper, we employ depth priors as the geometric representation. Our approach focuses on the integration of depth priors into various LLE frameworks using a unified methodology. This methodology comprises two key novel modules. First, a depth-aware feature extraction module is designed to inject depth priors into the image representation. Then, Hierarchical Depth-Guided Feature Fusion Module (HDGFFM) is formulated with a cross-domain attention mechanism, which combines depth-aware features with the original image features within the LLE model. We conducted extensive experiments on public low-light image and video enhancement benchmarks. The results illustrate that our designed framework significantly enhances existing LLE methods.
Authors: Jason Hu, Bowen Song, Xiaojian Xu, Liyue Shen, Jeffrey A. Fessler
Abstract: Diffusion models can learn strong image priors from underlying data distribution and use them to solve inverse problems, but the training process is computationally expensive and requires lots of data. Such bottlenecks prevent most existing works from being feasible for high-dimensional and high-resolution data such as 3D images. This paper proposes a method to learn an efficient data prior for the entire image by training diffusion models only on patches of images. Specifically, we propose a patch-based position-aware diffusion inverse solver, called PaDIS, where we obtain the score function of the whole image through scores of patches and their positional encoding and utilize this as the prior for solving inverse problems. First of all, we show that this diffusion model achieves an improved memory efficiency and data efficiency while still maintaining the capability to generate entire images via positional encoding. Additionally, the proposed PaDIS model is highly flexible and can be plugged in with different diffusion inverse solvers (DIS). We demonstrate that the proposed PaDIS approach enables solving various inverse problems in both natural and medical image domains, including CT reconstruction, deblurring, and superresolution, given only patch-based priors. Notably, PaDIS outperforms previous DIS methods trained on entire image priors in the case of limited training data, demonstrating the data efficiency of our proposed approach by learning patch-based prior.
Authors: Hanyuan Xiao, Yingshu Chen, Huajian Huang, Haolin Xiong, Jing Yang, Pratusha Prasad, Yajie Zhao
Abstract: Recent text-guided generation of individual 3D object has achieved great success using diffusion priors. However, these methods are not suitable for object insertion and replacement tasks as they do not consider the background, leading to illumination mismatches within the environment. To bridge the gap, we introduce an illumination-aware 3D scene editing pipeline for 3D Gaussian Splatting (3DGS) representation. Our key observation is that inpainting by the state-of-the-art conditional 2D diffusion model is consistent with background in lighting. To leverage the prior knowledge from the well-trained diffusion models for 3D object generation, our approach employs a coarse-to-fine objection optimization pipeline with inpainted views. In the first coarse step, we achieve image-to-3D lifting given an ideal inpainted view. The process employs 3D-aware diffusion prior from a view-conditioned diffusion model, which preserves illumination present in the conditioning image. To acquire an ideal inpainted image, we introduce an Anchor View Proposal (AVP) algorithm to find a single view that best represents the scene illumination in target region. In the second Texture Enhancement step, we introduce a novel Depth-guided Inpainting Score Distillation Sampling (DI-SDS), which enhances geometry and texture details with the inpainting diffusion prior, beyond the scope of the 3D-aware diffusion prior knowledge in the first coarse step. DI-SDS not only provides fine-grained texture enhancement, but also urges optimization to respect scene lighting. Our approach efficiently achieves local editing with global illumination consistency without explicitly modeling light transport. We demonstrate robustness of our method by evaluating editing in real scenes containing explicit highlight and shadows, and compare against the state-of-the-art text-to-3D editing methods.
Authors: Manuel Serna-Aguilera, Xuan Bac Nguyen, Han-Seok Seo, Khoa Luu
Abstract: Facial expressions and actions differ among different individuals at varying degrees of intensity given responses to external stimuli, particularly among those that are neurodivergent. Such behaviors affect people in terms of overall health, communication, and sensory processing. Deep learning can be responsibly leveraged to improve productivity in addressing this task, and help medical professionals to accurately understand such behaviors. In this work, we introduce the Video ASD dataset-a dataset that contains video frame convolutional and attention map feature data-to foster further progress in the task of ASD classification. Unlike many recent studies in ASD classification with MRI data, which require expensive specialized equipment, our method utilizes a powerful but relatively affordable GPU, a standard computer setup, and a video camera for inference. Results show that our model effectively generalizes and understands key differences in the distinct movements of the children. Additionally, we test foundation models on this data to showcase how movement noise affects performance and the need for more data and more complex labels.
Authors: Jidong Kuang, Hongsong Wang, Chaolei Han, Yang Zhang, Jie Gui
Abstract: Zero-shot action recognition, which addresses the issue of scalability and generalization in action recognition and allows the models to adapt to new and unseen actions dynamically, is an important research topic in computer vision communities. The key to zero-shot action recognition lies in aligning visual features with semantic vectors representing action categories. Most existing methods either directly project visual features onto the semantic space of text category or learn a shared embedding space between the two modalities. However, a direct projection cannot accurately align the two modalities, and learning robust and discriminative embedding space between visual and text representations is often difficult. To address these issues, we introduce Dual Visual-Text Alignment (DVTA) for skeleton-based zero-shot action recognition. The DVTA consists of two alignment modules--Direct Alignment (DA) and Augmented Alignment (AA)--along with a designed Semantic Description Enhancement (SDE). The DA module maps the skeleton features to the semantic space through a specially designed visual projector, followed by the SDE, which is based on cross-attention to enhance the connection between skeleton and text, thereby reducing the gap between modalities. The AA module further strengthens the learning of the embedding space by utilizing deep metric learning to learn the similarity between skeleton and text. Our approach achieves state-of-the-art performances on several popular zero-shot skeleton-based action recognition benchmarks. The code is available at: https://github.com/jidongkuang/DVTA.
Authors: Alexander Becker, Jan D. Wegner, Evans Dawoe, Konrad Schindler, William J. Thompson, Christian Bunn, Rachael D. Garrett, Fabio Castro-Llanos, Simon P. Hart, Wilma J. Blaser-Hart
Abstract: Reconciling agricultural production with climate-change mitigation is a formidable sustainability problem. Retaining trees in agricultural systems is one proposed solution, but the magnitude of the current and future-potential benefit that trees contribute to climate-change mitigation remains uncertain. Here, we help to resolve these issues across a West African region that produces ~60% of the world's cocoa, a crop contributing one of the highest carbon footprints of all foods. Using machine learning, we mapped shade-tree cover and carbon stocks across the region and found that existing average cover is low (~13%) and poorly aligned with climate threats. Yet, increasing shade-tree cover to a minimum of 30% could sequester an additional 307 million tonnes of CO2e, enough to offset ~167% of contemporary cocoa-related emissions in Ghana and C\^ote d'Ivoire--without reducing production. Our approach is transferable to other shade-grown crops and aligns with emerging carbon market and sustainability reporting frameworks.
Authors: Oguzhan Yigit, Richard C. Wilson
Abstract: The Laplace-Beltrami operator has established itself in the field of non-rigid shape analysis due to its many useful properties such as being invariant under isometric transformation, having a countable eigensystem forming an orthornormal basis, and fully characterizing geodesic distances of the manifold. However, this invariancy only applies under isometric deformations, which leads to a performance breakdown in many real-world applications. In recent years emphasis has been placed upon extracting optimal features using deep learning methods,however spectral signatures play a crucial role and still add value. In this paper we take a step back, revisiting the LBO and proposing a supervised way to learn several operators on a manifold. Depending on the task, by applying these functions, we can train the LBO eigenbasis to be more task-specific. The optimization of the LBO leads to enormous improvements to established descriptors such as the heat kernel signature in various tasks such as retrieval, classification, segmentation, and correspondence, proving the adaption of the LBO eigenbasis to both global and highly local learning settings.
Authors: Xiaobin Deng, Changyu Diao, Min Li, Ruohan Yu, Duanqing Xu
Abstract: 3D Gaussian Splatting (3DGS) has demonstrated outstanding performance in novel view synthesis, achieving a balance between rendering quality and real-time performance. 3DGS employs Adaptive Density Control (ADC) to increase the number of Gaussians. However, the clone and split operations within ADC are not sufficiently efficient, impacting optimization speed and detail recovery. Additionally, overfitted Gaussians that affect rendering quality may exist, and the original ADC is unable to remove them. To address these issues, we propose two key innovations: (1) Long-Axis Split, which precisely controls the position, shape, and opacity of child Gaussians to minimize the difference before and after splitting. (2) Recovery-Aware Pruning, which leverages differences in recovery speed after resetting opacity to prune overfitted Gaussians, thereby improving generalization performance. Experimental results show that our method significantly enhances rendering quality. Due to resubmission reasons, this version has been abandoned. The improved version is available at https://xiaobin2001.github.io/improved-gs-web .
Authors: Xiangyong Lu, Masanori Suganuma, Takayuki Okatani
Abstract: In real-world applications of image recognition tasks, such as human pose estimation, cameras often capture objects, like human bodies, at low resolutions. This scenario poses a challenge in extracting and leveraging multi-scale features, which is often essential for precise inference. To address this challenge, we propose a new attention mechanism, named cascaded multi-scale attention (CMSA), tailored for use in CNN-ViT hybrid architectures, to handle low-resolution inputs effectively. The design of CMSA enables the extraction and seamless integration of features across various scales without necessitating the downsampling of the input image or feature maps. This is achieved through a novel combination of grouped multi-head self-attention mechanisms with window-based local attention and cascaded fusion of multi-scale features over different scales. This architecture allows for the effective handling of features across different scales, enhancing the model's ability to perform tasks such as human pose estimation, head pose estimation, and more with low-resolution images. Our experimental results show that the proposed method outperforms existing state-of-the-art methods in these areas with fewer parameters, showcasing its potential for broad application in real-world scenarios where capturing high-resolution images is not feasible. Code is available at https://github.com/xyongLu/CMSA.
Authors: Bohan Li, Xin Jin, Jianan Wang, Yukai Shi, Yasheng Sun, Xiaofeng Wang, Zhuang Ma, Baao Xie, Chao Ma, Xiaokang Yang, Wenjun Zeng
Abstract: Recent diffusion models have demonstrated remarkable performance in both 3D scene generation and perception tasks. Nevertheless, existing methods typically separate these two processes, acting as a data augmenter to generate synthetic data for downstream perception tasks. In this work, we propose OccScene, a novel mutual learning paradigm that integrates fine-grained 3D perception and high-quality generation in a unified framework, achieving a cross-task win-win effect. OccScene generates new and consistent 3D realistic scenes only depending on text prompts, guided with semantic occupancy in a joint-training diffusion framework. To align the occupancy with the diffusion latent, a Mamba-based Dual Alignment module is introduced to incorporate fine-grained semantics and geometry as perception priors. Within OccScene, the perception module can be effectively improved with customized and diverse generated scenes, while the perception priors in return enhance the generation performance for mutual benefits. Extensive experiments show that OccScene achieves realistic 3D scene generation in broad indoor and outdoor scenarios, while concurrently boosting the perception models to achieve substantial performance improvements in the 3D perception task of semantic occupancy prediction.
Authors: Xiaojie Li, Jianlong Wu, Yue Yu, Liqiang Nie, Min Zhang
Abstract: Few-Shot Class-Incremental Learning (FSCIL) faces a critical challenge: balancing the retention of prior knowledge with the acquisition of new classes. Existing methods either freeze the backbone to prevent catastrophic forgetting, sacrificing plasticity, or add new modules, incurring high costs. These approaches treat pretrained models as black boxes, overlooking two key opportunities to exploit their internal capacity: reusing redundant representational space within layers and selectively adapting layers based on their sensitivity to forgetting. We propose CKPD-FSCIL, a unified framework that unlocks the underutilized capacity of pretrained weights, achieving a superior stability-plasticity balance with zero inference overhead. Our design integrates two continuously adapting mechanisms: At the weight level, a Continuous Knowledge-Preserving Decomposition mechanism uses feature covariance to split each weight matrix into a frozen subspace that safeguards prior knowledge and a learnable, redundant subspace for new tasks. At the layer level, a Continuous Adaptive Layer Selection mechanism leverages an Adapter Sensitivity Ratio to automatically select layers with the highest redundant capacity and lowest forgetting risk for adaptation. By targeting only safe, high-potential subspaces and layers, CKPD-FSCIL enables efficient adaptation. After each session, the learned adapters are merged back into the original weights, ensuring zero additional parameters or FLOPs during inference. Extensive experiments on multiple FSCIL benchmarks demonstrate that our method consistently outperforms state-of-the-art approaches in both adaptability and knowledge retention. The code is available at https://github.com/xiaojieli0903/CKPD-FSCIL.
Authors: Ketan Kotwal, Sebastien Marcel
Abstract: Demographic fairness in face recognition (FR) has emerged as a critical area of research, given its impact on fairness, equity, and reliability across diverse applications. As FR technologies are increasingly deployed globally, disparities in performance across demographic groups -- such as race, ethnicity, and gender -- have garnered significant attention. These biases not only compromise the credibility of FR systems but also raise ethical concerns, especially when these technologies are employed in sensitive domains. This review consolidates extensive research efforts providing a comprehensive overview of the multifaceted aspects of demographic fairness in FR. We systematically examine the primary causes, datasets, assessment metrics, and mitigation approaches associated with demographic disparities in FR. By categorizing key contributions in these areas, this work provides a structured approach to understanding and addressing the complexity of this issue. Finally, we highlight current advancements and identify emerging challenges that need further investigation. This article aims to provide researchers with a unified perspective on the state-of-the-art while emphasizing the critical need for equitable and trustworthy FR systems.
Authors: Hsiao-Yuan Chin, I-Chao Shen, Yi-Ting Chiu, Ariel Shamir, Bing-Yu Chen
Abstract: The ability to automatically complete a partial sketch that depicts a complex scene, e.g., "a woman chatting with a man in the park", is very useful. However, existing sketch generation methods create sketches from scratch; they do not complete a partial sketch in the style of the original. To address this challenge, we introduce AutoSketch, a styleaware vector sketch completion method that accommodates diverse sketch styles. Our key observation is that the style descriptions of a sketch in natural language preserve the style during automatic sketch completion. Thus, we use a pretrained vision-language model (VLM) to describe the styles of the partial sketches in natural language and replicate these styles using newly generated strokes. We initially optimize the strokes to match an input prompt augmented by style descriptions extracted from the VLM. Such descriptions allow the method to establish a diffusion prior in close alignment with that of the partial sketch. Next, we utilize the VLM to generate an executable style adjustment code that adjusts the strokes to conform to the desired style. We compare our method with existing methods across various sketch styles and prompts, performed extensive ablation studies and qualitative and quantitative evaluations, and demonstrate that AutoSketch can support various sketch scenarios.
Authors: Jindong Jiang, Xiuyu Li, Zhijian Liu, Muyang Li, Guo Chen, Zhiqi Li, De-An Huang, Guilin Liu, Zhiding Yu, Kurt Keutzer, Sungjin Ahn, Jan Kautz, Hongxu Yin, Yao Lu, Song Han, Wonmin Byeon
Abstract: Recent advances in video-based multimodal large language models (Video-LLMs) have significantly improved video understanding by processing videos as sequences of image frames. However, many existing methods treat frames independently in the vision backbone, lacking explicit temporal modeling, which limits their ability to capture dynamic patterns and efficiently handle long videos. To address these limitations, we introduce STORM (Spatiotemporal TOken Reduction for Multimodal LLMs), a novel architecture incorporating a dedicated temporal encoder between the image encoder and the LLM. Our temporal encoder leverages the Mamba State Space Model to integrate temporal information into image tokens, generating enriched representations that preserve inter-frame dynamics across the entire video sequence. This enriched encoding not only enhances video reasoning capabilities but also enables effective token reduction strategies, including test-time sampling and training-based temporal and spatial pooling, substantially reducing computational demands on the LLM without sacrificing key temporal information. By integrating these techniques, our approach simultaneously reduces training and inference latency while improving performance, enabling efficient and robust video understanding over extended temporal contexts. Extensive evaluations show that STORM achieves state-of-the-art results across various long video understanding benchmarks (more than 5% improvement on MLVU and LongVideoBench) while reducing the computation costs by up to $8\times$ and the decoding latency by 2.4-2.9$\times$ for the fixed numbers of input frames. Project page is available at https://research.nvidia.com/labs/lpr/storm
Authors: Cl\'ement Chadebec, Onur Tasar, Sanjeev Sreetharan, Benjamin Aubin
Abstract: In this paper, we introduce Latent Bridge Matching (LBM), a new, versatile and scalable method that relies on Bridge Matching in a latent space to achieve fast image-to-image translation. We show that the method can reach state-of-the-art results for various image-to-image tasks using only a single inference step. In addition to its efficiency, we also demonstrate the versatility of the method across different image translation tasks such as object removal, normal and depth estimation, and object relighting. We also derive a conditional framework of LBM and demonstrate its effectiveness by tackling the tasks of controllable image relighting and shadow generation. We provide an implementation at https://github.com/gojasper/LBM.
Authors: Khoi Do, Binh-Son Hua
Abstract: Score distillation sampling is an effective technique to generate 3D models from text prompts, utilizing pre-trained large-scale text-to-image diffusion models as guidance. However, the produced 3D assets tend to be over-saturating, over-smoothing, with limited diversity. These issues are results from a reverse Kullback-Leibler (KL) divergence objective, which makes the optimization unstable and results in mode-seeking behavior. In this paper, we derive a bounded score distillation objective based on Jensen-Shannon divergence (JSD), which stabilizes the optimization process and produces high-quality 3D generation. JSD can match well generated and target distribution, therefore mitigating mode seeking. We provide a practical implementation of JSD by utilizing the theory of generative adversarial networks to define an approximate objective function for the generator, assuming the discriminator is well trained. By assuming the discriminator following a log-odds classifier, we propose a minority sampling algorithm to estimate the gradients of our proposed objective, providing a practical implementation for JSD. We conduct both theoretical and empirical studies to validate our method. Experimental results on T3Bench demonstrate that our method can produce high-quality and diversified 3D assets.
Authors: Hiroyasu Akada, Jian Wang, Vladislav Golyanik, Christian Theobalt
Abstract: Egocentric 3D human pose estimation has been actively studied using cameras installed in front of a head-mounted device (HMD). While frontal placement is the optimal and the only option for some tasks, such as hand tracking, it remains unclear if the same holds for full-body tracking due to self-occlusion and limited field-of-view coverage. Notably, even the state-of-the-art methods often fail to estimate accurate 3D poses in many scenarios, such as when HMD users tilt their heads upward -- a common motion in human activities. A key limitation of existing HMD designs is their neglect of the back of the body, despite its potential to provide crucial 3D reconstruction cues. Hence, this paper investigates the usefulness of rear cameras for full-body tracking. We also show that simply adding rear views to the frontal inputs is not optimal for existing methods due to their dependence on individual 2D joint detectors without effective multi-view integration. To address this issue, we propose a new transformer-based method that refines 2D joint heatmap estimation with multi-view information and heatmap uncertainty, thereby improving 3D pose tracking. Also, we introduce two new large-scale datasets, Ego4View-Syn and Ego4View-RW, for a rear-view evaluation. Our experiments show that the new camera configurations with back views provide superior support for 3D pose tracking compared to only frontal placements. The proposed method achieves significant improvement over the current state of the art (>10% on MPJPE). The source code, trained models, and datasets are available on our project page at https://4dqv.mpi-inf.mpg.de/EgoRear/.
Authors: Fanhu Zeng, Hao Tang, Yihua Shao, Siyu Chen, Ling Shao, Yan Wang
Abstract: A high-performance image compression algorithm is crucial for real-time information transmission across numerous fields. Despite rapid progress in image compression, computational inefficiency and poor redundancy modeling still pose significant bottlenecks, limiting practical applications. Inspired by the effectiveness of state space models (SSMs) in capturing long-range dependencies, we leverage SSMs to address computational inefficiency in existing methods and improve image compression from multiple perspectives. In this paper, we integrate the advantages of SSMs for better efficiency-performance trade-off and propose an enhanced image compression approach through refined context modeling, which we term MambaIC. Specifically, we explore context modeling to adaptively refine the representation of hidden states. Additionally, we introduce window-based local attention into channel-spatial entropy modeling to reduce potential spatial redundancy during compression, thereby increasing efficiency. Comprehensive qualitative and quantitative results validate the effectiveness and efficiency of our approach, particularly for high-resolution image compression. Code is released at https://github.com/AuroraZengfh/MambaIC.
Authors: Hui Chen, Liangyu Liu, Xianchao Xiu, Wanquan Liu
Abstract: Hyperspectral unmixing (HU) is a critical yet challenging task in remote sensing. However, existing nonnegative matrix factorization (NMF) methods with graph learning mostly focus on first-order or second-order nearest neighbor relationships and usually require manual parameter tuning, which fails to characterize intrinsic data structures. To address the above issues, we propose a novel adaptive multi-order graph regularized NMF method (MOGNMF) with three key features. First, multi-order graph regularization is introduced into the NMF framework to exploit global and local information comprehensively. Second, these parameters associated with the multi-order graph are learned adaptively through a data-driven approach. Third, dual sparsity is embedded to obtain better robustness, i.e., $\ell_{1/2}$-norm on the abundance matrix and $\ell_{2,1}$-norm on the noise matrix. To solve the proposed model, we develop an alternating minimization algorithm whose subproblems have explicit solutions, thus ensuring effectiveness. Experiments on simulated and real hyperspectral data indicate that the proposed method delivers better unmixing results.
Authors: Haosheng Chen, Lian Luo, Mengjingcheng Mo, Zhanjie Wu, Guobao Xiao, Ji Gan, Jiaxu Leng, Xinbo Gao
Abstract: Event cameras, with microsecond temporal resolution and high dynamic range (HDR) characteristics, emit high-speed event stream for perception tasks. Despite the recent advancement in GNN-based perception methods, they are prone to use straightforward pairwise connectivity mechanisms in the pure Euclidean space where they struggle to capture long-range dependencies and fail to effectively characterize the inherent hierarchical structures of non-uniformly distributed event stream. To this end, in this paper we propose a novel approach named EHGCN, which is a pioneer to perceive event stream in both Euclidean and hyperbolic spaces for event vision. In EHGCN, we introduce an adaptive sampling strategy to dynamically regulate sampling rates, retaining discriminative events while attenuating chaotic noise. Then we present a Markov Vector Field (MVF)-driven motion-aware hyperedge generation method based on motion state transition probabilities, thereby eliminating cross-target spurious associations and providing critically topological priors while capturing long-range dependencies between events. Finally, we propose a Euclidean-Hyperbolic GCN to fuse the information locally aggregated and globally hierarchically modeled in Euclidean and hyperbolic spaces, respectively, to achieve hybrid event perception. Experimental results on event perception tasks such as object detection and recognition validate the effectiveness of our approach.
Authors: Oussema Dhaouadi, Johannes Meier, Luca Wahl, Jacques Kaiser, Luca Scalerandi, Nick Wandelburg, Zhuolun Zhou, Nijanthan Berinpanathan, Holger Banzhaf, Daniel Cremers
Abstract: Accurate 3D trajectory data is crucial for advancing autonomous driving. Yet, traditional datasets are usually captured by fixed sensors mounted on a car and are susceptible to occlusion. Additionally, such an approach can precisely reconstruct the dynamic environment in the close vicinity of the measurement vehicle only, while neglecting objects that are further away. In this paper, we introduce the DeepScenario Open 3D Dataset (DSC3D), a high-quality, occlusion-free dataset of 6 degrees of freedom bounding box trajectories acquired through a novel monocular camera drone tracking pipeline. Our dataset includes more than 175,000 trajectories of 14 types of traffic participants and significantly exceeds existing datasets in terms of diversity and scale, containing many unprecedented scenarios such as complex vehicle-pedestrian interaction on highly populated urban streets and comprehensive parking maneuvers from entry to exit. DSC3D dataset was captured in five various locations in Europe and the United States and include: a parking lot, a crowded inner-city, a steep urban intersection, a federal highway, and a suburban intersection. Our 3D trajectory dataset aims to enhance autonomous driving systems by providing detailed environmental 3D representations, which could lead to improved obstacle interactions and safety. We demonstrate its utility across multiple applications including motion prediction, motion planning, scenario mining, and generative reactive traffic agents. Our interactive online visualization platform and the complete dataset are publicly available at https://app.deepscenario.com, facilitating research in motion prediction, behavior modeling, and safety validation.
Authors: Yanshu Li, Jianjiang Yang, Ziteng Yang, Bozheng Li, Hongyang He, Zhengtao Yao, Ligong Han, Yingjie Victor Chen, Songlin Fei, Dongfang Liu, Ruixiang Tang
Abstract: Multimodal in-context learning (ICL) is emerging as a key capability that enables large vision-language models (LVLMs) to adapt to novel tasks without parameter updates, expanding their utility across various real-world applications. However, ICL remains unstable, even with well-matched in-context demonstrations (ICDs), suggesting that LVLMs struggle to fully utilize the provided context. While existing efforts focus on prompt engineering or post-hoc logit calibration, we instead investigate the underlying attention dynamics to overcome LVLMs' inherent limitations. We identify two critical deficits in their self-attention that impair effective ICL. To bridge the gap, we propose \textbf{Context-Aware Modulated Attention} (CAMA), a plug-and-play and training-free method that dynamically modulates LVLM's attention logits based on the input in-context sequence. CAMA employs a two-stage attention modulation to address both identified deficits, enhancing the focus on semantically significant tokens, particularly visual ones. Across four LVLMs and seven benchmarks, CAMA consistently outperforms vanilla models and baselines, demonstrating great effectiveness and generalization. It can also activate the desired effects of prompt engineering methods and remains robust under diverse sequence configurations. Thus, CAMA paves the way for deeper explorations of attention dynamics to advance multimodal reasoning.
Authors: Shenghui Chen, Po-han Li, Sandeep Chinchali, Ufuk Topcu
Abstract: Many decision-making tasks, where both accuracy and efficiency matter, still require human supervision. For example, tasks like traffic officers reviewing hour-long dashcam footage or researchers screening conference videos can benefit from concise summaries that reduce cognitive load and save time. Yet current vision-language models (VLMs) often produce verbose, redundant outputs that hinder task performance. Existing video caption evaluation depends on costly human annotations and overlooks the summaries' utility in downstream tasks. We address these gaps with Video-to-text Information Bottleneck Evaluation (VIBE), an annotation-free method that scores VLM outputs using two metrics: grounding (how well the summary aligns with visual content) and utility (how informative it is for the task). VIBE selects from randomly sampled VLM outputs by ranking them according to the two scores to support effective human decision-making. Human studies on LearningPaper24, SUTD-TrafficQA, and LongVideoBench show that summaries selected by VIBE consistently improve performance-boosting task accuracy by up to 61.23% and reducing response time by 75.77% compared to naive VLM summaries or raw video.
Authors: Xiao Yang, Jiyao Wang, Yuxuan Fan, Can Liu, Houcheng Su, Weichen Guo, Zitong Yu, Dengbo He, Kaishun Wu
Abstract: Remote physiological measurement (RPM) has emerged as a promising non-invasive method for monitoring physiological signals using the non-contact device. Although various domain adaptation and generalization methods were proposed to promote the adaptability of deep-based RPM models in unseen deployment environments, considerations in aspects such as privacy concerns and real-time adaptation restrict their application in real-world deployment. Thus, we aim to propose a novel fully Test-Time Adaptation (TTA) strategy tailored for RPM tasks in this work. Specifically, based on prior knowledge in physiology and our observations, we noticed not only there is spatio-temporal consistency in the frequency domain of BVP signals, but also that inconsistency in the time domain was significant. Given this, by leveraging both consistency and inconsistency priors, we introduce an innovative expert knowledge-based self-supervised \textbf{C}onsistency-\textbf{i}n\textbf{C}onsistency-\textbf{i}ntegration (\textbf{CiCi}) framework to enhances model adaptation during inference. Besides, our approach further incorporates a gradient dynamic control mechanism to mitigate potential conflicts between priors, ensuring stable adaptation across instances. Through extensive experiments on five diverse datasets under the TTA protocol, our method consistently outperforms existing techniques, presenting state-of-the-art performance in real-time self-supervised adaptation without accessing source data. The code will be released later.
Authors: Xinyu Chen, Haotian Zhai, Can Zhang, Xiupeng Shi, Ruirui Li
Abstract: In zero-shot setting, test-time adaptation adjusts pre-trained models using unlabeled data from the test phase to enhance performance on unknown test distributions. Existing cache-enhanced TTA methods rely on a low-entropy criterion to select samples for prototype construction, assuming intra-class compactness. However, low-entropy samples may be unreliable under distribution shifts, and the resulting prototypes may not ensure compact intra-class distributions. This study identifies a positive correlation between cache-enhanced performance and intra-class compactness. Based on this observation, we propose a Multi-Cache enhanced Prototype-based Test-Time Adaptation (MCP) featuring three caches: an entropy cache for initializing prototype representations with low-entropy samples, an align cache for integrating visual and textual information to achieve compact intra-class distributions, and a negative cache for prediction calibration using high-entropy samples. We further developed MCP++, a framework incorporating cross-modal prototype alignment and residual learning, introducing prototype residual fine-tuning. Comparative and ablation experiments across 15 downstream tasks demonstrate that the proposed method and framework achieve state-of-the-art generalization performance. Project Page available at: https://zhaihaotian.github.io/MCP-ICCV25/
Authors: Yuhang Ma, Yunhao Shui, Xiaoshi Wu, Keqiang Sun, Hongsheng Li
Abstract: Evaluating text-to-image generation models requires alignment with human perception, yet existing human-centric metrics are constrained by limited data coverage, suboptimal feature extraction, and inefficient loss functions. To address these challenges, we introduce Human Preference Score v3 (HPSv3). (1) We release HPDv3, the first wide-spectrum human preference dataset integrating 1.08M text-image pairs and 1.17M annotated pairwise comparisons from state-of-the-art generative models and low to high-quality real-world images. (2) We introduce a VLM-based preference model trained using an uncertainty-aware ranking loss for fine-grained ranking. Besides, we propose Chain-of-Human-Preference (CoHP), an iterative image refinement method that enhances quality without extra data, using HPSv3 to select the best image at each step. Extensive experiments demonstrate that HPSv3 serves as a robust metric for wide-spectrum image evaluation, and CoHP offers an efficient and human-aligned approach to improve image generation quality. The code and dataset are available at the HPSv3 Homepage.
Authors: Haoyang Li, Jiaqing Li, Jialun Cao, Zongyuan Yang, Yongping Xiong
Abstract: Large foundation models have achieved significant performance gains through scalable training on massive datasets. However, the field of \textbf{H}andwritten \textbf{M}athematical \textbf{E}xpression \textbf{R}ecognition (HMER) has been impeded by the scarcity of data, primarily due to the arduous and costly process of manual annotation. To bridge this gap, we propose a novel method integrating limited handwritten formulas with large-scale LaTeX-rendered formulas by developing a scalable data engine to generate complex and consistent LaTeX sequences. With this engine, we built the largest formula dataset to date, termed \texttt{Tex80M}, comprising over 80 million high-quality training instances. Then we propose \texttt{TexTeller}, the first HMER model trained at scale, by mix-training \texttt{Tex80M} with a relatively small HME dataset. The expansive training dataset and our refined pipeline have equipped \texttt{TexTeller} with state-of-the-art (SOTA) performance across nearly all benchmarks. To advance the field, we will openly release our complete model, entire dataset, and full codebase, enabling further research building upon our contributions.
Authors: Anushka Bhatt
Abstract: This study proposes a real-time system that translates voluntary eye blinks into Morse code, enabling communication for individuals with severe motor impairments. Using a standard webcam and computer vision, the system detects and classifies blinks as short (dot) or long (dash), then decodes them into alphanumeric characters. Experiments with five participants show 62% decoding accuracy and 18-20 seconds response times, demonstrating a viable, low-cost assistive communication method.
Authors: Shuting He, Peilin Ji, Yitong Yang, Changshuo Wang, Jiayi Ji, Yinglin Wang, Henghui Ding
Abstract: 3D Gaussian Splatting (3DGS) has recently emerged as a powerful alternative to Neural Radiance Fields (NeRF) for 3D scene representation, offering high-fidelity photorealistic rendering with real-time performance. Beyond novel view synthesis, the explicit and compact nature of 3DGS enables a wide range of downstream applications that require geometric and semantic understanding. This survey provides a comprehensive overview of recent progress in 3DGS applications. It first introduces 2D foundation models that support semantic understanding and control in 3DGS applications, followed by a review of NeRF-based methods that inform their 3DGS counterparts. We then categorize 3DGS applications into segmentation, editing, generation, and other functional tasks. For each, we summarize representative methods, supervision strategies, and learning paradigms, highlighting shared design principles and emerging trends. Commonly used datasets and evaluation protocols are also summarized, along with comparative analyses of recent methods across public benchmarks. To support ongoing research and development, a continually updated repository of papers, code, and resources is maintained at https://github.com/heshuting555/Awesome-3DGS-Applications.
URLs: https://github.com/heshuting555/Awesome-3DGS-Applications.
Authors: Yechan Kim, Dongho Yoon, Younkwan Lee, Unse Fatima, Hong Kook Kim, Songjae Lee, Sanga Park, Jeong Ho Park, Seonjong Kang, Moongu Jeon
Abstract: While previous studies on image segmentation focus on handling severe (or explicit) label noise, real-world datasets also exhibit subtle (or implicit) label imperfections. These arise from inherent challenges, such as ambiguous object boundaries and annotator variability. Although not explicitly present, such mild and latent noise can still impair model performance. Typical data augmentation methods, which apply identical transformations to the image and its label, risk amplifying these subtle imperfections and limiting the model's generalization capacity. In this paper, we introduce NSegment+, a novel augmentation framework that decouples image and label transformations to address such realistic noise for semantic segmentation. By introducing controlled elastic deformations only to segmentation labels while preserving the original images, our method encourages models to focus on learning robust representations of object structures despite minor label inconsistencies. Extensive experiments demonstrate that NSegment+ consistently improves performance, achieving mIoU gains of up to +2.29, +2.38, +1.75, and +3.39 in average on Vaihingen, LoveDA, Cityscapes, and PASCAL VOC, respectively-even without bells and whistles, highlighting the importance of addressing implicit label noise. These gains can be further amplified when combined with other training tricks, including CutMix and Label Smoothing.
Authors: Taeheon Kim, San Kim, Minhyuk Seo, Dongjae Jeon, Wonje Jeung, Jonghyun Choi
Abstract: Class-incremental with repetition (CIR), where previously trained classes repeatedly introduced in future tasks, is a more realistic scenario than the traditional class incremental setup, which assumes that each task contains unseen classes. CIR assumes that we can easily access abundant unlabeled data from external sources, such as the Internet. Therefore, we propose two components that efficiently use the unlabeled data to ensure the high stability and the plasticity of models trained in CIR setup. First, we introduce multi-level knowledge distillation (MLKD) that distills knowledge from multiple previous models across multiple perspectives, including features and logits, so the model can maintain much various previous knowledge. Moreover, we implement dynamic self-supervised loss (SSL) to utilize the unlabeled data that accelerates the learning of new classes, while dynamic weighting of SSL keeps the focus of training to the primary task. Both of our proposed components significantly improve the performance in CIR setup, achieving 2nd place in the CVPR 5th CLVISION Challenge.
Authors: Sukhun Ko, Dahyeon Kye, Kyle Min, Chanho Eom, Jihyong Oh
Abstract: Implicit Neural Representations (INRs) leverage neural networks to map coordinates to corresponding signals, enabling continuous and compact representations. This paradigm has driven significant advances in various vision tasks. However, existing INRs lack frequency selectivity, spatial localization, and sparse representations, leading to an over-reliance on redundant signal components. Consequently, they exhibit spectral bias, tending to learn low-frequency components early while struggling to capture fine high-frequency details. To address these issues, we propose FLAIR (Frequency- and Locality-Aware Implicit Neural Representations), which incorporates two key innovations. The first is RC-GAUSS, a novel activation designed for explicit frequency selection and spatial localization under the constraints of the time-frequency uncertainty principle (TFUP). The second is Wavelet-Energy-Guided Encoding (WEGE), which leverages the discrete wavelet transform (DWT) to compute energy scores and explicitly guide frequency information to the network. Our method consistently outperforms existing INRs in 2D image representation and restoration, as well as 3D reconstruction.
Authors: Andrea Atzori, Fadi Boutros, Naser Damer
Abstract: Face Image Quality Assessment (FIQA) aims to predict the utility of a face image for face recognition (FR) systems. State-of-the-art FIQA methods mainly rely on convolutional neural networks (CNNs), leaving the potential of Vision Transformer (ViT) architectures underexplored. This work proposes ViT-FIQA, a novel approach that extends standard ViT backbones, originally optimized for FR, through a learnable quality token designed to predict a scalar utility score for any given face image. The learnable quality token is concatenated with the standard image patch tokens, and the whole sequence is processed via global self-attention by the ViT encoders to aggregate contextual information across all patches. At the output of the backbone, ViT-FIQA branches into two heads: (1) the patch tokens are passed through a fully connected layer to learn discriminative face representations via a margin-penalty softmax loss, and (2) the quality token is fed into a regression head to learn to predict the face sample's utility. Extensive experiments on challenging benchmarks and several FR models, including both CNN- and ViT-based architectures, demonstrate that ViT-FIQA consistently achieves top-tier performance. These results underscore the effectiveness of transformer-based architectures in modeling face image utility and highlight the potential of ViTs as a scalable foundation for future FIQA research https://cutt.ly/irHlzXUC.
Authors: Gaston Gustavo Rios, Pedro Dal Bianco, Franco Ronchetti, Facundo Quiroga, Oscar Stanchi, Santiago Ponte Ah\'on, Waldo Hasperu\'e
Abstract: Sign Language Recognition (SLR) models face significant performance limitations due to insufficient training data availability. In this article, we address the challenge of limited data in SLR by introducing a novel and lightweight sign generation model based on CMLPe. This model, coupled with a synthetic data pretraining approach, consistently improves recognition accuracy, establishing new state-of-the-art results for the LSFB and DiSPLaY datasets using our Mamba-SL and Transformer-SL classifiers. Our findings reveal that synthetic data pretraining outperforms traditional augmentation methods in some cases and yields complementary benefits when implemented alongside them. Our approach democratizes sign generation and synthetic data pretraining for SLR by providing computationally efficient methods that achieve significant performance improvements across diverse datasets.
Authors: Youjia Zhang, Youngeun Kim, Young-Geun Choi, Hongyeob Kim, Huiling Liu, Sungeun Hong
Abstract: Test-time adaptation (TTA) enhances the zero-shot robustness under distribution shifts by leveraging unlabeled test data during inference. Despite notable advances, several challenges still limit its broader applicability. First, most methods rely on backpropagation or iterative optimization, which limits scalability and hinders real-time deployment. Second, they lack explicit modeling of class-conditional feature distributions. This modeling is crucial for producing reliable decision boundaries and calibrated predictions, but it remains underexplored due to the lack of both source data and supervision at test time. In this paper, we propose ADAPT, an Advanced Distribution-Aware and backPropagation-free Test-time adaptation method. We reframe TTA as a Gaussian probabilistic inference task by modeling class-conditional likelihoods using gradually updated class means and a shared covariance matrix. This enables closed-form, training-free inference. To correct potential likelihood bias, we introduce lightweight regularization guided by CLIP priors and a historical knowledge bank. ADAPT requires no source data, no gradient updates, and no full access to target data, supporting both online and transductive settings. Extensive experiments across diverse benchmarks demonstrate that our method achieves state-of-the-art performance under a wide range of distribution shifts with superior scalability and robustness.
Authors: Sumit Kumar Dam, Mrityunjoy Gain, Eui-Nam Huh, Choong Seon Hong
Abstract: Implicit Neural Representations (INRs) have emerged as a powerful alternative to traditional pixel-based formats by modeling images as continuous functions over spatial coordinates. A key challenge, however, lies in the spectral bias of neural networks, which tend to favor low-frequency components while struggling to capture high-frequency (HF) details such as sharp edges and fine textures. While prior approaches have addressed this limitation through architectural modifications or specialized activation functions, we propose an orthogonal direction by directly guiding the training process. Specifically, we introduce a two-stage training strategy where a neighbor-aware soft mask adaptively assigns higher weights to pixels with strong local variations, encouraging early focus on fine details. The model then transitions to full-image training. Experimental results show that our approach consistently improves reconstruction quality and complements existing INR methods. As a pioneering attempt to assign frequency-aware importance to pixels in image INR, our work offers a new avenue for mitigating the spectral bias problem.
Authors: Ziyang Yan, Ruikai Li, Zhiyong Cui, Bohan Li, Han Jiang, Yilong Ren, Aoyong Li, Zhenning Li, Sijia Wen, Haiyang Yu
Abstract: Online HD map construction is a fundamental task in autonomous driving systems, aiming to acquire semantic information of map elements around the ego vehicle based on real-time sensor inputs. Recently, several approaches have achieved promising results by incorporating offline priors such as SD maps and HD maps or by fusing multi-modal data. However, these methods depend on stale offline maps and multi-modal sensor suites, resulting in avoidable computational overhead at inference. To address these limitations, we employ a knowledge distillation strategy to transfer knowledge from multimodal models with prior knowledge to an efficient, low-cost, and vision-centric student model. Specifically, we propose MapKD, a novel multi-level cross-modal knowledge distillation framework with an innovative Teacher-Coach-Student (TCS) paradigm. This framework consists of: (1) a camera-LiDAR fusion model with SD/HD map priors serving as the teacher; (2) a vision-centric coach model with prior knowledge and simulated LiDAR to bridge the cross-modal knowledge transfer gap; and (3) a lightweight vision-based student model. Additionally, we introduce two targeted knowledge distillation strategies: Token-Guided 2D Patch Distillation (TGPD) for bird's eye view feature alignment and Masked Semantic Response Distillation (MSRD) for semantic learning guidance. Extensive experiments on the challenging nuScenes dataset demonstrate that MapKD improves the student model by +6.68 mIoU and +10.94 mAP while simultaneously accelerating inference speed. The code is available at:https://github.com/2004yan/MapKD2026.
Authors: Leonid Boytsov, Ameya Joshi, Filipe Condessa
Abstract: We experimented with front-end enhanced neural models where a differentiable and fully convolutional model with a skip connection is added before a frozen backbone classifier. By training such composite models using a small learning rate for about one epoch, we obtained models that retained the accuracy of the backbone classifier while being unusually resistant to gradient attacks-including APGD and FAB-T attacks from the AutoAttack package-which we attribute to gradient masking. Although gradient masking is not new, the degree we observe is striking for fully differentiable models without obvious gradient-shattering-e.g., JPEG compression-or gradient-diminishing components. The training recipe to produce such models is also remarkably stable and reproducible: We applied it to three datasets (CIFAR10, CIFAR100, and ImageNet) and several modern architectures (including vision Transformers) without a single failure case. While black-box attacks such as the SQUARE attack and zero-order PGD can partially overcome gradient masking, these attacks are easily defeated by simple randomized ensembles. We estimate that these ensembles achieve near-SOTA AutoAttack accuracy on CIFAR10, CIFAR100, and ImageNet (while retaining almost all clean accuracy of the original classifiers) despite having near-zero accuracy under adaptive attacks. Adversarially training the backbone further amplifies this front-end "robustness". On CIFAR10, the respective randomized ensemble achieved 90.8$\pm 2.5\%$ (99\% CI) accuracy under the full AutoAttack while having only 18.2$\pm 3.6\%$ accuracy under the adaptive attack ($\varepsilon=8/255$, $L^\infty$ norm). We conclude the paper with a discussion of whether randomized ensembling can serve as a practical defense. Code and instructions to reproduce key results are available. https://github.com/searchivarius/curious_case_of_gradient_masking
URLs: https://github.com/searchivarius/curious_case_of_gradient_masking
Authors: Buhua Liu, Shitong Shao, Bao Li, Lichen Bai, Zhiqiang Xu, Haoyi Xiong, James Kwok, Sumi Helal, Zeke Xie
Abstract: Diffusion models have emerged as the leading paradigm in generative modeling, excelling in various applications. Despite their success, these models often misalign with human intentions and generate results with undesired properties or even harmful content. Inspired by the success and popularity of alignment in tuning large language models, recent studies have investigated aligning diffusion models with human expectations and preferences. This work mainly reviews alignment of text-to-image diffusion models, covering advancements in fundamentals of alignment, alignment techniques of diffusion models, preference benchmarks, and evaluation for diffusion models. Moreover, we discuss key perspectives on current challenges and promising future directions on solving the remaining challenges in alignment of diffusion models. To the best of our knowledge, our work is the first comprehensive review paper for researchers and engineers to comprehend, practice, and research alignment of diffusion models.
Authors: Navya Sonal Agarwal, Jan Philipp Schneider, Kanchana Vaishnavi Gandikota, Syed Muhammad Kazim, John Meshreki, Ivo Ihrke, Michael Moeller
Abstract: The computational imaging technique of Fourier Ptychographic Microscopy (FPM) enables high-resolution imaging with a wide field of view and can serve as an extremely valuable tool, e.g. in the classification of cells in medical applications. However, reconstructing a high-resolution image from tens or even hundreds of measurements is computationally expensive, particularly for a wide field of view. Therefore, in this paper, we investigate the idea of classifying the image content in the FPM measurements directly without performing a reconstruction step first. We show that Convolutional Neural Networks (CNN) can extract meaningful information from measurement sequences, significantly outperforming the classification on a single band-limited image (up to 12 %) while being significantly more efficient than a reconstruction of a high-resolution image. Furthermore, we demonstrate that a learned multiplexing of several raw measurements allows maintaining the classification accuracy while reducing the amount of data (and consequently also the acquisition time) significantly.
Authors: Jianzhe Ma, Wenxuan Wang, Qin Jin
Abstract: Geometry problem solving, a crucial aspect of mathematical reasoning, is vital across various domains, including education, the assessment of AI's mathematical abilities, and multimodal capability evaluation. The recent surge in deep learning technologies, particularly the emergence of multimodal large language models, has significantly accelerated research in this area. This paper provides a survey of the applications of deep learning in geometry problem solving, including (i) a comprehensive summary of the relevant tasks in geometry problem solving; (ii) a thorough review of related deep learning methods; (iii) a detailed analysis of evaluation metrics and methods; and (iv) a critical discussion of the current challenges and future directions that can be explored. Our objective is to offer a comprehensive and practical reference of deep learning for geometry problem solving, thereby fostering further advancements in this field. We create a continuously updated list of papers on GitHub: https://github.com/majianz/dl4gps.
Authors: Aiden Ochoa, Xinyuan Xu, Xing Wang
Abstract: With ever-increasing data volumes, it is essential to develop automated approaches for identifying nanoscale defects in transmission electron microscopy (TEM) images. However, compared to features in conventional photographs, nanoscale defects in TEM images exhibit far greater variation due to the complex contrast mechanisms and intricate defect structures. These challenges often result in much less labeled data and higher rates of annotation errors, posing significant obstacles to improving machine learning model performance for TEM image analysis. To address these limitations, we examined transfer learning by leveraging large, pre-trained models used for natural images. We demonstrated that by using the pre-trained encoder and L2-regularization, semantically complex features are ignored in favor of simpler, more reliable cues, substantially improving the model performance. However, this improvement cannot be captured by conventional evaluation metrics such as F1-score, which can be skewed by human annotation errors treated as ground truth. Instead, we introduced novel evaluation metrics that are independent of the annotation accuracy. Using grain boundary detection in UO2 TEM images as a case study, we found that our approach led to a 57% increase in defect detection rate, which is a robust and holistic measure of model performance on the TEM dataset used in this work. Finally, we showed that model self-confidence is only achieved through transfer learning and fine-tuning of very deep layers.
Authors: Jianqiang Xiao, Yuexuan Sun, Yixin Shao, Boxi Gan, Rongqiang Liu, Yanjing Wu, Weili Guan, Xiang Deng
Abstract: Aerial navigation is a fundamental yet underexplored capability in embodied intelligence, enabling agents to operate in large-scale, unstructured environments where traditional navigation paradigms fall short. However, most existing research follows the Vision-and-Language Navigation (VLN) paradigm, which heavily depends on sequential linguistic instructions, limiting its scalability and autonomy. To address this gap, we introduce UAV-ON, a benchmark for large-scale Object Goal Navigation (ObjectNav) by aerial agents in open-world environments, where agents operate based on high-level semantic goals without relying on detailed instructional guidance as in VLN. UAV-ON comprises 14 high-fidelity Unreal Engine environments with diverse semantic regions and complex spatial layouts, covering urban, natural, and mixed-use settings. It defines 1270 annotated target objects, each characterized by an instance-level instruction that encodes category, physical footprint, and visual descriptors, allowing grounded reasoning. These instructions serve as semantic goals, introducing realistic ambiguity and complex reasoning challenges for aerial agents. To evaluate the benchmark, we implement several baseline methods, including Aerial ObjectNav Agent (AOA), a modular policy that integrates instruction semantics with egocentric observations for long-horizon, goal-directed exploration. Empirical results show that all baselines struggle in this setting, highlighting the compounded challenges of aerial navigation and semantic goal grounding. UAV-ON aims to advance research on scalable UAV autonomy driven by semantic goal descriptions in complex real-world environments.
Authors: Seungyeon Kim, Junsu Ha, Young Hun Kim, Yonghyeon Lee, Frank C. Park
Abstract: Articulated object recognition -- the task of identifying both the geometry and kinematic joints of objects with movable parts -- is essential for enabling robots to interact with everyday objects such as doors and laptops. However, existing approaches often rely on strong assumptions, such as a known number of articulated parts; require additional inputs, such as depth images; or involve complex intermediate steps that can introduce potential errors -- limiting their practicality in real-world settings. In this paper, we introduce ScrewSplat, a simple end-to-end method that operates solely on RGB observations. Our approach begins by randomly initializing screw axes, which are then iteratively optimized to recover the object's underlying kinematic structure. By integrating with Gaussian Splatting, we simultaneously reconstruct the 3D geometry and segment the object into rigid, movable parts. We demonstrate that our method achieves state-of-the-art recognition accuracy across a diverse set of articulated objects, and further enables zero-shot, text-guided manipulation using the recovered kinematic model. See the project website at: https://screwsplat.github.io.
Authors: Pengwei Sun, Wei Peng, Lun Yu Li, Yixin Wang, Kilian M. Pohl
Abstract: Counterfactual generation offers a principled framework for simulating hypothetical changes in medical imaging, with potential applications in understanding disease mechanisms and generating physiologically plausible data. However, generating realistic structural 3D brain MRIs that respect anatomical and causal constraints remains challenging due to data scarcity, structural complexity, and the lack of standardized evaluation protocols. In this work, we convert six generative models into 3D counterfactual approaches by incorporating an anatomy-guided framework based on a causal graph, in which regional brain volumes serve as direct conditioning inputs. Each model is evaluated with respect to composition, reversibility, realism, effectiveness and minimality on T1-weighted brain MRIs (T1w MRIs) from the Alzheimer's Disease Neuroimaging Initiative (ADNI). In addition, we test the generalizability of each model with respect to T1w MRIs of the National Consortium on Alcohol and Neurodevelopment in Adolescence (NCANDA). Our results indicate that anatomically grounded conditioning successfully modifies the targeted anatomical regions; however, it exhibits limitations in preserving non-targeted structures. Beyond laying the groundwork for more interpretable and clinically relevant generative modeling of brain MRIs, this benchmark highlights the need for novel architectures that more accurately capture anatomical interdependencies.
Authors: Gideon N. L. Rouwendaal, Dani\"el Boeke, Inge L. Cox, Henk G. van der Poel, Margriet C. van Dijk-de Haan, Regina G. H. Beets-Tan, Thierry N. Boellaard, Wilson Silva
Abstract: Accurate preoperative prediction of erectile dysfunction (ED) is important for counseling patients undergoing radical prostatectomy. While clinical features are established predictors, the added value of preoperative MRI remains underexplored. We investigate whether MRI provides additional predictive value for ED at 12 months post-surgery, evaluating four modeling strategies: (1) a clinical-only baseline, representing current state-of-the-art; (2) classical models using handcrafted anatomical features derived from MRI; (3) deep learning models trained directly on MRI slices; and (4) multimodal fusion of imaging and clinical inputs. Imaging-based models (maximum AUC 0.569) slightly outperformed handcrafted anatomical approaches (AUC 0.554) but fell short of the clinical baseline (AUC 0.663). Fusion models offered marginal gains (AUC 0.586) but did not exceed clinical-only performance. SHAP analysis confirmed that clinical features contributed most to predictive performance. Saliency maps from the best-performing imaging model suggested a predominant focus on anatomically plausible regions, such as the prostate and neurovascular bundles. While MRI-based models did not improve predictive performance over clinical features, our findings suggest that they try to capture patterns in relevant anatomical structures and may complement clinical predictors in future multimodal approaches.
Authors: Luca Zedda, Andrea Loddo, Cecilia Di Ruberto, Carsten Marr
Abstract: Red blood cells (RBCs) are essential to human health, and their precise morphological analysis is important for diagnosing hematological disorders. Despite the promise of foundation models in medical diagnostics, comprehensive AI solutions for RBC analysis remain scarce. We present RedDino, a self-supervised foundation model designed for RBC image analysis. RedDino uses an RBC-specific adaptation of the DINOv2 self-supervised learning framework and is trained on a curated dataset of 1.25 million RBC images from diverse acquisition modalities and sources. Extensive evaluations show that RedDino outperforms existing state-of-the-art models on RBC shape classification. Through assessments including linear probing and nearest neighbor classification, we confirm its strong feature representations and generalization ability. Our main contributions are: (1) a foundation model tailored for RBC analysis, (2) ablation studies exploring DINOv2 configurations for RBC modeling, and (3) a detailed evaluation of generalization performance. RedDino addresses key challenges in computational hematology by capturing nuanced morphological features, advancing the development of reliable diagnostic tools. The source code and pretrained models for RedDino are available at https://github.com/Snarci/RedDino, and the pretrained models can be downloaded from our Hugging Face collection at https://huggingface.co/collections/Snarcy/reddino-689a13e29241d2e5690202fc
URLs: https://github.com/Snarci/RedDino,, https://huggingface.co/collections/Snarcy/reddino-689a13e29241d2e5690202fc
Authors: Bingquan Dai, Li Ray Luo, Qihong Tang, Jie Wang, Xinyu Lian, Hao Xu, Minghan Qin, Xudong Xu, Bo Dai, Haoqian Wang, Zhaoyang Lyu, Jiangmiao Pang
Abstract: Reconstructing 3D objects into editable programs is pivotal for applications like reverse engineering and shape editing. However, existing methods often rely on limited domain-specific languages (DSLs) and small-scale datasets, restricting their ability to model complex geometries and structures. To address these challenges, we introduce MeshCoder, a novel framework that reconstructs complex 3D objects from point clouds into editable Blender Python scripts. We develop a comprehensive set of expressive Blender Python APIs capable of synthesizing intricate geometries. Leveraging these APIs, we construct a large-scale paired object-code dataset, where the code for each object is decomposed into distinct semantic parts. Subsequently, we train a multimodal large language model (LLM) that translates 3D point cloud into executable Blender Python scripts. Our approach not only achieves superior performance in shape-to-code reconstruction tasks but also facilitates intuitive geometric and topological editing through convenient code modifications. Furthermore, our code-based representation enhances the reasoning capabilities of LLMs in 3D shape understanding tasks. Together, these contributions establish MeshCoder as a powerful and flexible solution for programmatic 3D shape reconstruction and understanding. The project homepage is available at \href{https://daibingquan.github.io/MeshCoder}{this link}.