Peer Revenue Play credits. Real papers.

Papers

Fresh from arXiv, typeset properly. Highlighted words trade on the market.

Computer Science Computer Vision

cs.CV Watch Computer Vision (sign in to use)

Thursday 6 November

Computer Vision

Thinking with Video: Video Generation as a Promising Multimodal Reasoning Paradigm

The "Thinking with Text" and "Thinking with Images" paradigms significantly improve the reasoning abilities of large language models (LLMs) and Vision-Language Models (VLMs). However, these paradigms have inherent limitations. (1) Images capture only single moments and fail to represent dynamic processes or continuous changes, and (2) The separation of text and vision as distinct modalities, which hinders unified multimodal understanding and generation. Therefore, we propose "Thinking with Video", a new paradigm that leverages video generation models such as Sora-2 to use video frames as a unified medium for multimodal reasoning. To support this exploration, we developed the Video Thinking Benchmark (VideoThinkBench), which covers both vision-centric tasks (e.g., Eyeballing Puzzles) and text-centric tasks (e.g., GSM8K and MMMU). Our evaluation on VideoThinkBench establishes Sora-2 as a capable reasoner. On vision-centric tasks, Sora-2 is comparable to state-of-the-art (SOTA) VLMs, and even surpasses GPT-5 by 10% on eyeballing puzzles. On text-centric tasks, Sora-2 achieves 92% accuracy on MATH, and 69.2% accuracy on MMMU. Furthermore, we systematically analyze the source of these abilities. We also find that self-consistency and in-context learning can improve Sora-2's performance. In summary, our findings show that the video generation model is the potential unified multimodal understanding and generation model, positioning "Thinking with Video" as a potential unified multimodal reasoning paradigm.
Read the full abstract

All authors Jingqi Tong, Yurong Mou, Hangcheng Li, Mingzhe Li, Yongzhuo Yang, Ming Zhang, Qiguang Chen, Tianyi Liang, Xiaomeng Hu, Yining Zheng, Xinchi Chen, Jun Zhao, Xuanjing Huang, Xipeng Qiu

Computer Vision · Cross-listed from cs.RO

Evo-1: Lightweight Vision-Language-Action Model with Preserved Semantic Alignment

Vision-Language-Action (VLA) models have emerged as a powerful framework that unifies perception, language, and control, enabling robots to perform diverse tasks through multimodal understanding. However, current VLA models typically contain massive parameters and rely heavily on large-scale robot data pretraining, leading to high computational costs during training, as well as limited deployability for real-time inference. Moreover, most training paradigms often degrade the perceptual representations of the vision-language backbone, resulting in overfitting and poor generalization to downstream tasks. In this work, we present Evo-1, a lightweight VLA model that reduces computation and improves deployment efficiency, while maintaining strong performance without pretraining on robot data. Evo-1 builds on a native multimodal Vision-Language model (VLM), incorporating a novel cross-modulated diffusion transformer along with an optimized integration module, together forming an effective architecture. We further introduce a two-stage training paradigm that progressively aligns action with perception, preserving the representations of the VLM. Notably, with only 0.77 billion parameters, Evo-1 achieves state-of-the-art results on the Meta-World and RoboTwin suite, surpassing the previous best models by 12.4% and 6.9%, respectively, and also attains a competitive result of 94.8% on LIBERO. In real-world evaluations, Evo-1 attains a 78% success rate with high inference frequency and low memory overhead, outperforming all baseline methods. We release code, data, and model weights to facilitate future research on lightweight and efficient VLA models.
Read the full abstract

All authors Tao Lin, Yilei Zhong, Yuxin Du, Jingjing Zhang, Jiting Liu, Yinxinyu Chen, Encheng Gu, Ziyan Liu, Hongyi Cai, Yanwen Zou, Lixing Zou, Zhaoye Zhou, Gen Li, Bo Zhao

Computer Vision

Learning from Single Timestamps: Complexity Estimation in Laparoscopic Cholecystectomy

Purpose: Accurate assessment of surgical complexity is essential in Laparoscopic Cholecystectomy (LC), where severe inflammation is associated with longer operative times and increased risk of postoperative complications. The Parkland Grading Scale (PGS) provides a clinically validated framework for stratifying inflammation severity; however, its automation in surgical videos remains largely unexplored, particularly in realistic scenarios where complete videos must be analyzed without prior manual curation. Methods: In this work, we introduce STC-Net, a novel framework for SingleTimestamp-based Complexity estimation in LC via the PGS, designed to operate under weak temporal supervision. Unlike prior methods limited to static images or manually trimmed clips, STC-Net operates directly on full videos. It jointly performs temporal localization and grading through a localization, window proposal, and grading module. We introduce a novel loss formulation combining hard and soft localization objectives and background-aware grading supervision. Results: Evaluated on a private dataset of 1,859 LC videos, STC-Net achieves an accuracy of 62.11% and an F1-score of 61.42%, outperforming non-localized baselines by over 10% in both metrics and highlighting the effectiveness of weak supervision for surgical complexity assessment. Conclusion: STC-Net demonstrates a scalable and effective approach for automated PGS-based surgical complexity estimation from full LC videos, making it promising for post-operative analysis and surgical training.
Read the full abstract

All authors Dimitrios Anastasiou, Santiago Barbarisi, Lucy Culshaw, Jayna Patel, Evangelos B. Mazomenos, Imanol Luengo, Danail Stoyanov

Computer Vision

THEval. Evaluation Framework for Talking Head Video Generation

Video generation has achieved remarkable progress, with generated videos increasingly resembling real ones. However, the rapid advance in generation has outpaced the development of adequate evaluation metrics. Currently, the assessment of talking head generation primarily relies on limited metrics, evaluating general video quality, lip synchronization, and on conducting user studies. Motivated by this, we propose a new evaluation framework comprising 8 metrics related to three dimensions (i) quality, (ii) naturalness, and (iii) synchronization. In selecting the metrics, we place emphasis on efficiency, as well as alignment with human preferences. Based on this considerations, we streamline to analyze fine-grained dynamics of head, mouth, and eyebrows, as well as face quality. Our extensive experiments on 85,000 videos generated by 17 state-of-the-art models suggest that while many algorithms excel in lip synchronization, they face challenges with generating expressiveness and artifact-free details. These videos were generated based on a novel real dataset, that we have curated, in order to mitigate bias of training data. Our proposed benchmark framework is aimed at evaluating the improvement of generative methods. Original code, dataset and leaderboards will be publicly released and regularly updated with new methods, in order to reflect progress in the field.
Read the full abstract

All authors Nabyl Quignon, Baptiste Chopin, Yaohui Wang, Antitza Dantcheva

Computer Vision · Cross-listed from eess.IV

$\mu$NeuFMT: Optical-Property-Adaptive Fluorescence Molecular Tomography via Implicit Neural Representation

Fluorescence Molecular Tomography (FMT) is a promising technique for non-invasive 3D visualization of fluorescent probes, but its reconstruction remains challenging due to the inherent ill-posedness and reliance on inaccurate or often-unknown tissue optical properties. While deep learning methods have shown promise, their supervised nature limits generalization beyond training data. To address these problems, we propose $\mu$NeuFMT, a self-supervised FMT reconstruction framework that integrates implicit neural-based scene representation with explicit physical modeling of photon propagation. Its key innovation lies in jointly optimize both the fluorescence distribution and the optical properties ($\mu$) during reconstruction, eliminating the need for precise prior knowledge of tissue optics or pre-conditioned training data. We demonstrate that $\mu$NeuFMT robustly recovers accurate fluorophore distributions and optical coefficients even with severely erroneous initial values (0.5$\times$ to 2$\times$ of ground truth). Extensive numerical, phantom, and in vivo validations show that $\mu$NeuFMT outperforms conventional and supervised deep learning approaches across diverse heterogeneous scenarios. Our work establishes a new paradigm for robust and accurate FMT reconstruction, paving the way for more reliable molecular imaging in complex clinically related scenarios, such as fluorescence guided surgery.
Read the full abstract

All authors Shihan Zhao, Jianru Zhang, Yanan Wu, Linlin Li, Siyuan Shen, Xingjun Zhu, Guoyan Zheng, Jiahua Jiang, Wuwei Ren

Computer Vision · Cross-listed from cs.LG

Distribution-Aware Tensor Decomposition for Compression of Convolutional Neural Networks

Neural networks are widely used for image-related tasks but typically demand considerable computing power. Once a network has been trained, however, its memory- and compute-footprint can be reduced by compression. In this work, we focus on compression through tensorization and low-rank representations. Whereas classical approaches search for a low-rank approximation by minimizing an isotropic norm such as the Frobenius norm in weight-space, we use data-informed norms that measure the error in function space. Concretely, we minimize the change in the layer's output distribution, which can be expressed as $\lVert (W - \widetilde{W}) \Sigma^{1/2}\rVert_F$ where $\Sigma^{1/2}$ is the square root of the covariance matrix of the layer's input and $W$, $\widetilde{W}$ are the original and compressed weights. We propose new alternating least square algorithms for the two most common tensor decompositions (Tucker-2 and CPD) that directly optimize the new norm. Unlike conventional compression pipelines, which almost always require post-compression fine-tuning, our data-informed approach often achieves competitive accuracy without any fine-tuning. We further show that the same covariance-based norm can be transferred from one dataset to another with only a minor accuracy drop, enabling compression even when the original training dataset is unavailable. Experiments on several CNN architectures (ResNet-18/50, and GoogLeNet) and datasets (ImageNet, FGVC-Aircraft, Cifar10, and Cifar100) confirm the advantages of the proposed method.
Read the full abstract

All authors Alper Kalle, Theo Rudkiewicz, Mohamed-Oumar Ouerfelli, Mohamed Tamaazousti

Computer Vision

Landslide Hazard Mapping with Geospatial Foundation Models: Geographical Generalizability, Data Scarcity, and Band Adaptability

Landslides cause severe damage to lives, infrastructure, and the environment, making accurate and timely mapping essential for disaster preparedness and response. However, conventional deep learning models often struggle when applied across different sensors, regions, or under conditions of limited training data. To address these challenges, we present a three-axis analytical framework of sensor, label, and domain for adapting geospatial foundation models (GeoFMs), focusing on Prithvi-EO-2.0 for landslide mapping. Through a series of experiments, we show that it consistently outperforms task-specific CNNs (U-Net, U-Net++), vision transformers (Segformer, SwinV2-B), and other GeoFMs (TerraMind, SatMAE). The model, built on global pretraining, self-supervision, and adaptable fine-tuning, proved resilient to spectral variation, maintained accuracy under label scarcity, and generalized more reliably across diverse datasets and geographic settings. Alongside these strengths, we also highlight remaining challenges such as computational cost and the limited availability of reusable AI-ready training data for landslide research. Overall, our study positions GeoFMs as a step toward more robust and scalable approaches for landslide risk reduction and environmental monitoring.
Read the full abstract

All authors Wenwen Li, Sizhe Wang, Hyunho Lee, Chenyan Lu, Sujit Roy, Rahul Ramachandran, Chia-Yu Hsu

Computer Vision

V-Thinker: Interactive Thinking with Images

Empowering Large Multimodal Models (LMMs) to deeply integrate image interaction with long-horizon reasoning capabilities remains a long-standing challenge in this field. Recent advances in vision-centric reasoning explore a promising "Thinking with Images" paradigm for LMMs, marking a shift from image-assisted reasoning to image-interactive thinking. While this milestone enables models to focus on fine-grained image regions, progress remains constrained by limited visual tool spaces and task-specific workflow designs. To bridge this gap, we present V-Thinker, a general-purpose multimodal reasoning assistant that enables interactive, vision-centric thinking through end-to-end reinforcement learning. V-Thinker comprises two key components: (1) a Data Evolution Flywheel that automatically synthesizes, evolves, and verifies interactive reasoning datasets across three dimensions-diversity, quality, and difficulty; and (2) a Visual Progressive Training Curriculum that first aligns perception via point-level supervision, then integrates interactive reasoning through a two-stage reinforcement learning framework. Furthermore, we introduce VTBench, an expert-verified benchmark targeting vision-centric interactive reasoning tasks. Extensive experiments demonstrate that V-Thinker consistently outperforms strong LMM-based baselines in both general and interactive reasoning scenarios, providing valuable insights for advancing image-interactive reasoning applications.
Read the full abstract

All authors Runqi Qiao, Qiuna Tan, Minghan Yang, Guanting Dong, Peiqing Yang, Shiqiang Lang, Enhui Wan, Xiaowan Wang, Yida Xu, Lan Yang, Chong Sun, Chen Li, Jing Lyu, Honggang Zhang

Computer Vision

Solving Convex Partition Visual Jigsaw Puzzles

Jigsaw puzzle solving requires the rearrangement of unordered pieces into their original pose in order to reconstruct a coherent whole, often an image, and is known to be an intractable problem. While the possible impact of automatic puzzle solvers can be disruptive in various application domains, most of the literature has focused on developing solvers for square jigsaw puzzles, severely limiting their practical use. In this work, we significantly expand the types of puzzles handled computationally, focusing on what is known as Convex Partitions, a major subset of polygonal puzzles whose pieces are convex. We utilize both geometrical and pictorial compatibilities, introduce a greedy solver, and report several performance measures next to the first benchmark dataset of such puzzles.
Read the full abstract

Computer Vision

HideAndSeg: an AI-based tool with automated prompting for octopus segmentation in natural habitats

Analyzing octopuses in their natural habitats is challenging due to their camouflage capability, rapid changes in skin texture and color, non-rigid body deformations, and frequent occlusions, all of which are compounded by variable underwater lighting and turbidity. Addressing the lack of large-scale annotated datasets, this paper introduces HideAndSeg, a novel, minimally supervised AI-based tool for segmenting videos of octopuses. It establishes a quantitative baseline for this task. HideAndSeg integrates SAM2 with a custom-trained YOLOv11 object detector. First, the user provides point coordinates to generate the initial segmentation masks with SAM2. These masks serve as training data for the YOLO model. After that, our approach fully automates the pipeline by providing a bounding box prompt to SAM2, eliminating the need for further manual intervention. We introduce two unsupervised metrics - temporal consistency $DICE_t$ and new component count $NC_t$ - to quantitatively evaluate segmentation quality and guide mask refinement in the absence of ground-truth data, i.e., real-world information that serves to train, validate, and test AI models. Results show that HideAndSeg achieves satisfactory performance, reducing segmentation noise compared to the manually prompted approach. Our method can re-identify and segment the octopus even after periods of complete occlusion in natural environments, a scenario in which the manually prompted model fails. By reducing the need for manual analysis in real-world scenarios, this work provides a practical tool that paves the way for more efficient behavioral studies of wild cephalopods.
Read the full abstract

All authors Alan de Aguiar, Michaella Pereira Andrade, Charles Morphy D. Santos, João Paulo Gois

Computer Vision · Cross-listed from cs.LG

On the Equivalence of Regression and Classification

A formal link between regression and classification has been tenuous. Even though the margin maximization term $\|w\|$ is used in support vector regression, it has at best been justified as a regularizer. We show that a regression problem with $M$ samples lying on a hyperplane has a one-to-one equivalence with a linearly separable classification task with $2M$ samples. We show that margin maximization on the equivalent classification task leads to a different regression formulation than traditionally used. Using the equivalence, we demonstrate a “regressability” measure, that can be used to estimate the difficulty of regressing a dataset, without needing to first learn a model for it. We use the equivalence to train neural networks to learn a linearizing map, that transforms input variables into a space where a linear regressor is adequate.
Read the full abstract

All authors Jayadeva, Naman Dwivedi, Hari Krishnan, N.M. Anoop Krishnan

Computer Vision

DORAEMON: A Unified Library for Visual Object Modeling and Representation Learning at Scale

DORAEMON is an open-source PyTorch library that unifies visual object modeling and representation learning across diverse scales. A single YAML-driven workflow covers classification, retrieval and metric learning; more than 1000 pretrained backbones are exposed through a timm-compatible interface, together with modular losses, augmentations and distributed-training utilities. Reproducible recipes match or exceed reference results on ImageNet-1K, MS-Celeb-1M and Stanford online products, while one-command export to ONNX or HuggingFace bridges research and deployment. By consolidating datasets, models, and training techniques into one platform, DORAEMON offers a scalable foundation for rapid experimentation in visual recognition and representation learning, enabling efficient transfer of research advances to real-world applications. The repository is available at https://github.com/wuji3/DORAEMON.
Read the full abstract

All authors Ke Du, Yimin Peng, Chao Gao, Fan Zhou, Siqiao Xue

Computer Vision

BoRe-Depth: Self-supervised Monocular Depth Estimation with Boundary Refinement for Embedded Systems

Depth estimation is one of the key technologies for realizing 3D perception in unmanned systems. Monocular depth estimation has been widely researched because of its low-cost advantage, but the existing methods face the challenges of poor depth estimation performance and blurred object boundaries on embedded systems. In this paper, we propose a novel monocular depth estimation model, BoRe-Depth, which contains only 8.7M parameters. It can accurately estimate depth maps on embedded systems and significantly improves boundary quality. Firstly, we design an Enhanced Feature Adaptive Fusion Module (EFAF) which adaptively fuses depth features to enhance boundary detail representation. Secondly, we integrate semantic knowledge into the encoder to improve the object recognition and boundary perception capabilities. Finally, BoRe-Depth is deployed on NVIDIA Jetson Orin, and runs efficiently at 50.7 FPS. We demonstrate that the proposed model significantly outperforms previous lightweight models on multiple challenging datasets, and we provide detailed ablation studies for the proposed methods. The code is available at https://github.com/liangxiansheng093/BoRe-Depth.
Read the full abstract

All authors Chang Liu, Juan Li, Sheng Zhang, Chang Liu, Jie Li, Xu Zhang

Computer Vision

Multi-Task Learning for Visually Grounded Reasoning in Gastrointestinal VQA

We present a multi-task framework for the MediaEval Medico 2025 challenge, leveraging a LoRA-tuned Florence-2 model for simultaneous visual question answering (VQA), explanation generation, and visual grounding. The proposed system integrates three curated datasets: (1) Kvasir-VQA-x1 for question-answer learning, (2) a synthetically enriched explanation dataset offering structured medical reasoning, and (3) text-to-region pairs linking visual features with segmentation masks. This multi-task setup enables the model to jointly learn visual grounding, reasoning, and interpretation, producing responses that are both accurate and interpretable. Extensive evaluation demonstrates that our approach substantially improves over single-task baselines in both answer accuracy and visual localization, highlighting the effectiveness of grounded multi-task learning for medical VQA applications.
Read the full abstract

All authors Itbaan Safwan, Muhammad Annas Shaikh, Muhammad Haaris, Ramail Khan, Muhammad Atif Tahir

Computer Vision · Cross-listed from cs.RO

GraSP-VLA: Graph-based Symbolic Action Representation for Long-Horizon Planning with VLA Policies

Deploying autonomous robots that can learn new skills from demonstrations is an important challenge of modern robotics. Existing solutions often apply end-to-end imitation learning with Vision-Language Action (VLA) models or symbolic approaches with Action Model Learning (AML). On the one hand, current VLA models are limited by the lack of high-level symbolic planning, which hinders their abilities in long-horizon tasks. On the other hand, symbolic approaches in AML lack generalization and scalability perspectives. In this paper we present a new neuro-symbolic approach, GraSP-VLA, a framework that uses a Continuous Scene Graph representation to generate a symbolic representation of human demonstrations. This representation is used to generate new planning domains during inference and serves as an orchestrator for low-level VLA policies, scaling up the number of actions that can be reproduced in a row. Our results show that GraSP-VLA is effective for modeling symbolic representations on the task of automatic planning domain generation from observations. In addition, results on real-world experiments show the potential of our Continuous Scene Graph representation to orchestrate low-level VLA policies in long-horizon tasks.
Read the full abstract

All authors Maëlic Neau, Zoe Falomir, Paulo E. Santos, Anne-Gwenn Bosser, Cédric Buche

Computer Vision

A MATLAB tutorial on deep feature extraction combined with chemometrics for analytical applications

Background In analytical chemistry, spatial information about materials is commonly captured through imaging techniques, such as traditional color cameras or with advanced hyperspectral cameras and microscopes. However, efficiently extracting and analyzing this spatial information for exploratory and predictive purposes remains a challenge, especially when using traditional chemometric methods. Recent advances in deep learning and artificial intelligence have significantly enhanced image processing capabilities, enabling the extraction of multiscale deep features that are otherwise challenging to capture with conventional image processing techniques. Despite the wide availability of open-source deep learning models, adoption in analytical chemistry remains limited because of the absence of structured, step-by-step guidance for implementing these models. Results This tutorial aims to bridge this gap by providing a step-by-step guide for applying deep learning approaches to extract spatial information from imaging data and integrating it with other data sources, such as spectral information. Importantly, the focus of this work is not on training deep learning models for image processing but on using existing open source models to extract deep features from imaging data. Significance The tutorial provides MATLAB code tutorial demonstrations, showcasing the processing of imaging data from various imaging modalities commonly encountered in analytical chemistry. Readers must run the tutorial steps on their own datasets using the codes presented in this tutorial.
Read the full abstract

All authors Puneet Mishra, Martijntje Vollebregt, Yizhou Ma, Maria Font-i-Furnols

Computer Vision

Evaluating the Impact of Weather-Induced Sensor Occlusion on BEVFusion for 3D Object Detection

Accurate 3D object detection is essential for automated vehicles to navigate safely in complex real-world environments. Bird's Eye View (BEV) representations, which project multi-sensor data into a top-down spatial format, have emerged as a powerful approach for robust perception. Although BEV-based fusion architectures have demonstrated strong performance through multimodal integration, the effects of sensor occlusions, caused by environmental conditions such as fog, haze, or physical obstructions, on 3D detection accuracy remain underexplored. In this work, we investigate the impact of occlusions on both camera and Light Detection and Ranging (LiDAR) outputs using the BEVFusion architecture, evaluated on the nuScenes dataset. Detection performance is measured using mean Average Precision (mAP) and the nuScenes Detection Score (NDS). Our results show that moderate camera occlusions lead to a 41.3% drop in mAP (from 35.6% to 20.9%) when detection is based only on the camera. On the other hand, LiDAR sharply drops in performance only under heavy occlusion, with mAP falling by 47.3% (from 64.7% to 34.1%), with a severe impact on long-range detection. In fused settings, the effect depends on which sensor is occluded: occluding the camera leads to a minor 4.1% drop (from 68.5% to 65.7%), while occluding LiDAR results in a larger 26.8% drop (to 50.1%), revealing the model's stronger reliance on LiDAR for the task of 3D object detection. Our results highlight the need for future research into occlusion-aware evaluation methods and improved sensor fusion techniques that can maintain detection accuracy in the presence of partial sensor failure or degradation due to adverse environmental conditions.
Read the full abstract

All authors Sanjay Kumar, Tim Brophy, Eoin Martino Grua, Ganesh Sistu, Valentina Donzella, Ciaran Eising

Computer Vision

Comparative Study of CNN Architectures for Binary Classification of Horses and Motorcycles in the VOC 2008 Dataset

This paper presents a comprehensive evaluation of nine convolutional neural network architectures for binary classification of horses and motorcycles in the VOC 2008 dataset. We address the significant class imbalance problem by implementing minority-class augmentation techniques. Our experiments compare modern architectures including ResNet-50, ConvNeXt-Tiny, DenseNet-121, and Vision Transformer across multiple performance metrics. Results demonstrate substantial performance variations, with ConvNeXt-Tiny achieving the highest Average Precision (AP) of 95.53% for horse detection and 89.12% for motorcycle detection. We observe that data augmentation significantly improves minority class detection, particularly benefiting deeper architectures. This study provides insights into architecture selection for imbalanced binary classification tasks and quantifies the impact of data augmentation strategies in mitigating class imbalance issues in object detection.
Read the full abstract

Computer Vision

Submanifold Sparse Convolutional Networks for Automated 3D Segmentation of Kidneys and Kidney Tumours in Computed Tomography

Accurate delineation of kidney tumours in Computed Tomography (CT) is essential for downstream quantitative analysis and precision oncology, but manual segmentation is a specialised task, time-consuming and difficult to scale. Automated 3D segmentation remains challenging because CT scans are large volumetric images, making high-resolution dense convolutional networks computationally expensive and often dependent on downsampling or patch-based inference. We propose a two-stage 3D segmentation methodology based on voxel sparsification and submanifold sparse convolutional networks (SSCNs). Stage 1 uses a low-resolution sparse network to identify a region of interest (ROI); Stage 2 applies a high-resolution sparse network for refined segmentation within the cropped ROI. This enables native high-resolution 3D processing while reducing memory use and inference time. We evaluate the method on the KiTS23 renal cancer CT dataset using 5-fold cross-validation. Our method achieved Dice similarity coefficients of 95.8% for kidneys + masses, 85.7% for tumours + cysts, and 80.3% for tumours alone, competitive with top KiTS23 approaches. In direct comparisons on the same cross-validation folds, the proposed sparse method achieves tumour + cyst and tumour-only Dice scores comparable to, and slightly higher than, a patch-based nnU-Net baseline, while consistently requiring less VRAM and shorter inference time across the tested hardware. Across the tested GPUs, our sparse model is markedly faster than both nnU-Net and the zero-shot zoom-out/zoom-in foundation model SegVol, which localises kidneys well but underperforms on small heterogeneous lesions. Compared to an equivalent dense implementation of the same architecture, the proposed sparse approach achieves up to a 60% reduction in inference time and up to a 75% reduction in VRAM usage across both CPU and the GPU configurations tested.
Read the full abstract

All authors Saúl Alonso-Monsalve, Leigh H. Whitehead, Adam Aurisano, Lorena Escudero Sanchez

Computer Vision

RISE-T2V: Rephrasing and Injecting Semantics with LLM for Expansive Text-to-Video Generation

Most text-to-video(T2V) diffusion models depend on pre-trained text encoders for semantic alignment, yet they often fail to maintain video quality when provided with concise prompts rather than well-designed ones. The primary issue lies in their limited textual semantics understanding. Moreover, these text encoders cannot rephrase prompts online to better align with user intentions, which limits both the scalability and usability of the models, To address these challenges, we introduce RISE-T2V, which uniquely integrates the processes of prompt rephrasing and semantic feature extraction into a single and seamless step instead of two separate steps. RISE-T2V is universal and can be applied to various pre-trained LLMs and video diffusion models(VDMs), significantly enhancing their capabilities for T2V tasks. We propose an innovative module called the Rephrasing Adapter, enabling diffusion models to utilize text hidden states during the next token prediction of the LLM as a condition for video generation. By employing a Rephrasing Adapter, the video generation model can implicitly rephrase basic prompts into more comprehensive representations that better match the user's intent. Furthermore, we leverage the powerful capabilities of LLMs to enable video generation models to accomplish a broader range of T2V tasks. Extensive experiments demonstrate that RISE-T2V is a versatile framework applicable to different video diffusion model architectures, significantly enhancing the ability of T2V models to generate high-quality videos that align with user intent. Visual results are available on the webpage at https://rise-t2v.github.io.
Read the full abstract

All authors Xiangjun Zhang, Litong Gong, Yinglin Zheng, Yansong Liu, Wentao Jiang, Mingyi Xu, Biao Wang, Tiezheng Ge, Ming Zeng

Computer Vision

Deep learning-based object detection of offshore platforms on Sentinel-1 Imagery and the impact of synthetic training data

The recent and ongoing expansion of marine infrastructure, including offshore wind farms, oil and gas platforms, artificial islands, and aquaculture facilities, highlights the need for effective monitoring systems. The development of robust models for offshore infrastructure detection relies on comprehensive, balanced datasets, but falls short when samples are scarce, particularly for underrepresented object classes, shapes, and sizes. By training deep learning-based YOLOv10 object detection models with a combination of synthetic and real Sentinel-1 satellite imagery acquired in the fourth quarter of 2023 from four regions (Caspian Sea, South China Sea, Gulf of Guinea, and Coast of Brazil), this study investigates the use of synthetic training data to enhance model performance. We evaluated this approach by applying the model to detect offshore platforms in three unseen regions (Gulf of Mexico, North Sea, Persian Gulf) and thereby assess geographic transferability. This region-holdout evaluation demonstrated that the model generalises beyond the training areas. In total, 3,529 offshore platforms were detected, including 411 in the North Sea, 1,519 in the Gulf of Mexico, and 1,593 in the Persian Gulf. The model achieved an F1 score of 0.85, which improved to 0.90 upon incorporating synthetic data. We analysed how synthetic data enhances the representation of unbalanced classes and overall model performance, taking a first step toward globally transferable detection of offshore infrastructure. This study underscores the importance of balanced datasets and highlights synthetic data generation as an effective strategy to address common challenges in remote sensing, demonstrating the potential of deep learning for scalable, global offshore infrastructure monitoring.
Read the full abstract

Computer Vision

Vision Foundation Models in Agriculture: Toward Domain-Specific Adaptation for Weed Herbicide Trials Assessment

Herbicide field trials require accurate identification of plant species and assessment of herbicide-induced damage across diverse environments. While general-purpose vision foundation models have shown promising results in complex visual domains, their performance can be limited in agriculture, where fine-grained distinctions between species and damage types are critical. In this work, we adapt a general-purpose vision foundation model to herbicide trial characterization. Trained using a self-supervised learning approach on a large, curated agricultural dataset, the model learns rich and transferable representations optimized for herbicide trials images. Our domain-specific model significantly outperforms the best general-purpose foundation model in both species identification (F1 score improvement from 0.91 to 0.94) and damage classification (from 0.26 to 0.33). Under unseen conditions (new locations and other time), it achieves even greater gains (species identification from 0.56 to 0.66; damage classification from 0.17 to 0.27). In domain-shift scenarios, such as drone imagery, it maintains strong performance (species classification from 0.49 to 0.60). Additionally, we show that domain-specific pretraining enhances segmentation accuracy, particularly in low-annotation regimes. An annotation-efficiency analysis reveals that, under unseen conditions, the domain-specific model achieves 5.4% higher F1 score than the general-purpose model, while using 80% fewer labeled samples. These results demonstrate the generalization capabilities of domain-specific foundation models and their potential to significantly reduce manual annotation efforts, offering a scalable and automated solution for herbicide trial analysis.
Read the full abstract

All authors Leire Benito-Del-Valle, Artzai Picón, Daniel Mugica, Manuel Ramos, Eva Portillo, Javier Romero, Carlos Javier Jimenez, Ramón Navarra-Mestre

Computer Vision

FastGS: Training 3D Gaussian Splatting in 100 Seconds

The dominant 3D Gaussian splatting (3DGS) acceleration methods fail to properly regulate the number of Gaussians during training, causing redundant computational time overhead. In this paper, we propose FastGS, a novel, simple, and general acceleration framework that fully considers the importance of each Gaussian based on multi-view consistency, efficiently solving the trade-off between training time and rendering quality. We innovatively design a densification and pruning strategy based on multi-view consistency, dispensing with the budgeting mechanism. Extensive experiments on Mip-NeRF 360, Tanks & Temples, and Deep Blending datasets demonstrate that our method significantly outperforms the state-of-the-art methods in training speed, achieving a 3.32$\times$ training acceleration and comparable rendering quality compared with DashGaussian on the Mip-NeRF 360 dataset and a 15.45$\times$ acceleration compared with vanilla 3DGS on the Deep Blending dataset. We demonstrate that FastGS exhibits strong generality, delivering 2-7$\times$ training acceleration across various tasks, including dynamic scene reconstruction, surface reconstruction, sparse-view reconstruction, large-scale reconstruction, and simultaneous localization and mapping. The project page is available at https://fastgs.github.io/
Read the full abstract

All authors Shiwei Ren, Tianci Wen, Yongchun Fang, Biao Lu

Computer Vision

DINOv2 Driven Gait Representation Learning for Video-Based Visible-Infrared Person Re-identification

Video-based Visible-Infrared person re-identification (VVI-ReID) aims to retrieve the same pedestrian across visible and infrared modalities from video sequences. Existing methods tend to exploit modality-invariant visual features but largely overlook gait features, which are not only modality-invariant but also rich in temporal dynamics, thus limiting their ability to model the spatiotemporal consistency essential for cross-modal video matching. To address these challenges, we propose a DINOv2-Driven Gait Representation Learning (DinoGRL) framework that leverages the rich visual priors of DINOv2 to learn gait features complementary to appearance cues, facilitating robust sequence-level representations for cross-modal retrieval. Specifically, we introduce a Semantic-Aware Silhouette and Gait Learning (SASGL) model, which generates and enhances silhouette representations with general-purpose semantic priors from DINOv2 and jointly optimizes them with the ReID objective to achieve semantically enriched and task-adaptive gait feature learning. Furthermore, we develop a Progressive Bidirectional Multi-Granularity Enhancement (PBMGE) module, which progressively refines feature representations by enabling bidirectional interactions between gait and appearance streams across multiple spatial granularities, fully leveraging their complementarity to enhance global representations with rich local details and produce highly discriminative features. Extensive experiments on HITSZ-VCM and BUPT datasets demonstrate the superiority of our approach, significantly outperforming existing state-of-the-art methods.
Read the full abstract

All authors Yujie Yang, Shuang Li, Jun Ye, Neng Dong, Fan Li, Huafeng Li

Computer Vision

Proto-LeakNet: Towards Signal-Leak Aware Attribution in Synthetic Human Face Imagery

The growing sophistication of synthetic image and deepfake generation models has turned source attribution and authenticity verification into a critical challenge for modern computer vision systems. Recent studies suggest that diffusion pipelines unintentionally imprint persistent statistical traces, known as signal-leaks, within their outputs, particularly in latent representations. Building on this observation, we propose Proto-LeakNet, a signal-leak-aware and interpretable attribution framework that integrates Closed-set classification with a density-based Open-set evaluation on the learned embeddings, enabling analysis of unseen generators without retraining. Acting in the latent domain of diffusion models, our method re-simulates partial forward diffusion to expose residual generator-specific cues. A temporal attention encoder aggregates multi-step latent features, while a feature-weighted prototype head structures the embedding space and enables transparent attribution. Trained solely on closed data and achieving a Macro AUC of 98.13%, Proto-LeakNet learns a latent geometry that remains robust under post-processing, surpassing state-of-the-art methods, and achieves strong separability both between real images and known generators, and between known and unseen ones. The codebase is available at the following link: https://github.com/claudiunderthehood/Proto-LeakNet .
Read the full abstract

Computer Vision

MedSapiens: Taking a Pose to Rethink Medical Imaging Landmark Detection

This paper does not introduce a novel architecture; instead, it revisits a fundamental yet overlooked baseline: adapting human-centric foundation models for anatomical landmark detection in medical imaging. While landmark detection has traditionally relied on domain-specific models, the emergence of large-scale pre-trained vision models presents new opportunities. In this study, we investigate the adaptation of Sapiens, a human-centric foundation model designed for pose estimation, to medical imaging through multi-dataset pretraining, establishing a new state of the art across multiple datasets. Our proposed model, MedSapiens, demonstrates that human-centric foundation models, inherently optimized for spatial pose localization, provide strong priors for anatomical landmark detection, yet this potential has remained largely untapped. We benchmark MedSapiens against existing state-of-the-art models, achieving up to 5.26% improvement over generalist models and up to 21.81% improvement over specialist models in the average success detection rate (SDR). To further assess MedSapiens adaptability to novel downstream tasks with few annotations, we evaluate its performance in limited-data settings, achieving 2.69% improvement over the few-shot state of the art in SDR. Code and model weights are available at https://github.com/xmed-lab/MedSapiens .
Read the full abstract

All authors Marawan Elbatel, Anbang Wang, Keyuan Liu, Kaouther Mouheb, Enrique Almar-Munoz, Lizhuo Lin, Yanqi Yang, Karim Lekadir, Xiaomeng Li

Computer Vision

AStF: Motion Style Transfer via Adaptive Statistics Fusor

Human motion style transfer allows characters to appear less rigidity and more realism with specific style. Traditional arbitrary image style transfer typically process mean and variance which is proved effective. Meanwhile, similar methods have been adapted for motion style transfer. However, due to the fundamental differences between images and motion, relying on mean and variance is insufficient to fully capture the complex dynamic patterns and spatiotemporal coherence properties of motion data. Building upon this, our key insight is to bring two more coefficient, skewness and kurtosis, into the analysis of motion style. Specifically, we propose a novel Adaptive Statistics Fusor (AStF) which consists of Style Disentanglement Module (SDM) and High-Order Multi-Statistics Attention (HOS-Attn). We trained our AStF in conjunction with a Motion Consistency Regularization (MCR) discriminator. Experimental results show that, by providing a more comprehensive model of the spatiotemporal statistical patterns inherent in dynamic styles, our proposed AStF shows proficiency superiority in motion style transfers over state-of-the-arts. Our code and model are available at https://github.com/CHMimilanlan/AStF.
Read the full abstract

All authors Hanmo Chen, Chenghao Xu, Jiexi Yan, Cheng Deng

Computer Vision

Covariance Descriptors Meet General Vision Encoders: Riemannian Deep Learning for Medical Image Classification

Covariance descriptors capture second-order statistics of image features. They have shown strong performance in general computer vision tasks, but remain underexplored in medical imaging. We investigate their effectiveness for both conventional and learning-based medical image classification, with a particular focus on SPDNet, a classification network specifically designed for symmetric positive definite (SPD) matrices. We propose constructing covariance descriptors from features extracted by pre-trained general vision encoders (GVEs) and comparing them with handcrafted descriptors. Two GVEs - DINOv2 and MedSAM - are evaluated across eleven binary and multi-class datasets from the MedMNSIT benchmark. Our results show that covariance descriptors derived from GVE features consistently outperform those derived from handcrafted features. Moreover, SPDNet yields superior performance to state-of-the-art methods when combined with DINOv2 features. Our findings highlight the potential of combining covariance descriptors with powerful pretrained vision encoders for medical image analysis.
Read the full abstract

All authors Josef Mayr, Anna Reithmeir, Maxime Di Folco, Julia A. Schnabel

Computer Vision

Systematic Evaluation of Preprocessing Techniques for Accurate Image Registration in Digital Pathology

Image registration refers to the process of spatially aligning two or more images by mapping them into a common coordinate system, so that corresponding anatomical or tissue structures are matched across images. In digital pathology, registration enables direct comparison and integration of information from different stains or imaging modalities, sup-porting applications such as biomarker analysis and tissue reconstruction. Accurate registration of images from different modalities is an essential step in digital pathology. In this study, we investigated how various color transformation techniques affect image registration between hematoxylin and eosin (H&E) stained images and non-linear multimodal images. We used a dataset of 20 tissue sample pairs, with each pair undergoing several preprocessing steps, including different color transformation (CycleGAN, Macenko, Reinhard, Vahadane), inversion, contrast adjustment, intensity normalization, and denoising. All images were registered using the VALIS registration method, which first applies rigid registration and then performs non-rigid registration in two steps on both low and high-resolution images. Registration performance was evaluated using the relative Target Registration Error (rTRE). We reported the median of median rTRE values (MMrTRE) and the average of median rTRE values (AMrTRE) for each method. In addition, we performed a custom point-based evaluation using ten manually selected key points. Registration was done separately for two scenarios, using either the original or inverted multimodal images. In both scenarios, CycleGAN color transformation achieved the lowest registration errors, while the other methods showed higher errors. These findings show that applying color transformation before registration improves alignment between images from different modalities and supports more reliable analysis in digital pathology.
Read the full abstract

Computer Vision

Seeing Straight: Document Orientation Detection for Efficient OCR

Despite significant advances in document understanding, determining the correct orientation of scanned or photographed documents remains a critical pre-processing step in the real world settings. Accurate rotation correction is essential for enhancing the performance of downstream tasks such as Optical Character Recognition (OCR) where misalignment commonly arises due to user errors, particularly incorrect base orientations of the camera during capture. In this study, we first introduce OCR-Rotation-Bench (ORB), a new benchmark for evaluating OCR robustness to image rotations, comprising (i) ORB-En, built from rotation-transformed structured and free-form English OCR datasets, and (ii) ORB-Indic, a novel multilingual set spanning 11 Indic mid to low-resource languages. We also present a fast, robust and lightweight rotation classification pipeline built on the vision encoder of Phi-3.5-Vision model with dynamic image cropping, fine-tuned specifically for 4-class rotation task in a standalone fashion. Our method achieves near-perfect 96% and 92% accuracy on identifying the rotations respectively on both the datasets. Beyond classification, we demonstrate the critical role of our module in boosting OCR performance: closed-source (up to 14%) and open-weights models (up to 4x) in the simulated real-world setting.
Read the full abstract

All authors Suranjan Goswami, Abhinav Ravi, Raja Kolla, Ali Faraz, Shaharukh Khan, Akash, Chandra Khatri, Shubham Agarwal