CV

arXiv 2603.23376v1

cs.CV自动驾驶CV

YOLOv10 with Kolmogorov-Arnold networks and vision-language foundation models for interpretable object detection and trustworthy multimodal AI in computer vision perception

This paper presents an interpretable object detection framework using Kolmogorov-Arnold networks to enhance trustworthiness in autonomous vehicle perception systems. The approach addresses the critical limitation of limited transparency in confidence scores during visually degraded or ambiguous driving scenarios. A Kolmogorov-Arnold network serves as an interpretable post-hoc surrogate model for YOLOv10 detections, utilizing seven geometric and semantic features to assess detection reliability. The additive spline-based architecture enables direct visualization of feature contributions, revealing when confidence scores are well-supported versus unreliable. Experimental validation on COCO dataset and University of Bath campus images demonstrates accurate trustworthiness estimation for autonomous driving perception.

Unknown

arXiv 2603.23037v1

cs.CVCVTransformer

Grounding Sim-to-Real Generalization in Dexterous Manipulation: An Empirical Study with Vision-Language-Action Models

This paper investigates Sim-to-Real generalization for dexterous manipulation tasks using Vision-Language-Action (VLA) models. The study empirically examines key factors affecting transfer from simulation to real-world deployment, including multi-level domain randomization, photorealistic rendering, physics-realistic modeling, and reinforcement learning updates. A comprehensive evaluation protocol is designed to quantify real-world manipulation performance, providing insights for developing generalist robot control policies that can effectively bridge the simulation-to-reality gap in dexterous manipulation scenarios.

arXiv 2603.22876v1

cs.CV自动驾驶CV

Gau-Occ: Geometry-Completed Gaussians for Multi-Modal 3D Occupancy Prediction

This paper presents Gau-Occ, a multi-modal framework for 3D semantic occupancy prediction in autonomous driving that models scenes as compact collections of semantic 3D Gaussians, bypassing traditional dense volumetric processing. The proposed LiDAR Completion Diffuser recovers missing structures from sparse LiDAR point clouds to initialize robust Gaussian anchors, while Gaussian Anchor Fusion efficiently integrates multi-view image semantics through geometry-aligned 2D sampling and cross-modal alignment. By refining these compact Gaussian descriptors, the method achieves both spatial consistency and semantic discriminability for comprehensive 3D scene understanding in autonomous vehicles.

Chengxin Lv +3

arXiv 2603.22852v1

cs.CV大语言模型端到端

End-to-End Training for Unified Tokenization and Latent Denoising

Latent diffusion models enable high-fidelity synthesis by operating in learned latent spaces. However, current approaches require complex multi-stage training where tokenizers must be trained separately before diffusion models. This paper proposes UNITE, an autoencoder architecture that unifies tokenization and latent diffusion through a single Generative Encoder with shared weights. The key innovation is treating tokenization and generation as the same latent inference problem under different conditioning regimes. The method introduces a single-stage training procedure that jointly optimizes both tokenization and generation tasks via two forward passes, enabling gradients to jointly shape the latent space for improved visual representation learning.

UNITE Authors

arXiv 2603.22283v1

cs.CVCVTransformer

DualCoT-VLA: Visual-Linguistic Chain of Thought via Parallel Reasoning for Vision-Language-Action Models

Vision-Language-Action (VLA) models enable robots to map visual observations and language instructions directly to robotic actions. However, existing VLA models struggle with complex multi-step tasks requiring logical planning and precise manipulations. Current Chain-of-Thought approaches have limitations in simultaneously capturing low-level visual details and high-level logical planning, as well as suffering from high inference latency with compounding errors. This paper proposes DualCoT-VLA, a novel visual-linguistic CoT method with parallel reasoning mechanism that integrates visual CoT for comprehensive multi-modal reasoning, enabling robots to perform effective thinking before acting for manipulation tasks.

DualCoT-VLA Authors

arXiv 2603.22280v1

cs.CV大语言模型CV

The Dual Mechanisms of Spatial Reasoning in Vision-Language Models

This paper investigates how vision-language models associate objects with their properties and spatial relations in multimodal tasks like image captioning and visual question answering. The research reveals that VLMs employ two concurrent mechanisms for spatial reasoning: the language model backbone represents content-independent spatial relations on visual tokens, while the vision encoder encodes object layouts that are directly utilized by the language model. The dominant spatial information originates from the vision encoder and is distributed globally across visual tokens, extending beyond object regions. These findings provide insights into the internal workings of multimodal models for spatial understanding.

arXiv 2603.22278v1

cs.CVCV3D检测

UniDex: A Robot Foundation Suite for Universal Dexterous Hand Control from Egocentric Human Videos

This paper presents UniDex, a robot foundation suite addressing the challenges of dexterous manipulation by combining a large-scale robot-centric dataset with a unified vision-language-action (VLA) policy. The system transforms egocentric human videos into robot-executable trajectories across eight dexterous hand embodiments through a human-in-the-loop retargeting procedure. By operating on explicit 3D pointclouds with human hands masked, the approach narrows kinematic and visual gaps between human and robot domains. The introduced Function-Actuator-Aligned Space (FAAS) provides a unified action space for universal dexterous hand control.

UniDex Team

arXiv 2603.22264v1

cs.RO端到端CV

DexDrummer: In-Hand, Contact-Rich, and Long-Horizon Dexterous Robot Drumming

This paper addresses the challenging problem of in-hand, contact-rich, and long-horizon dexterous robot manipulation by proposing drumming as a comprehensive testbed. The DexDrummer framework employs a hierarchical object-centric bimanual policy that combines trajectory planning with residual reinforcement learning corrections, enabling effective sim-to-real transfer. The approach specifically targets dexterous manipulation skills including in-hand control for drumstick stabilization, contact-rich striking interactions, and long-horizon rhythmic coordination across multiple drums. By integrating these three challenging aspects into a single complex task, this work advances the field of robotic dexterity and manipulation planning.

DexDrummer Team

arXiv 2603.22263v1

cs.CV大语言模型CV

SpatialReward: Verifiable Spatial Reward Modeling for Fine-Grained Spatial Consistency in Text-to-Image Generation

This paper presents SpatialReward, a novel verifiable reward model designed to evaluate fine-grained spatial relationships in text-to-image generation. The approach employs a multi-stage pipeline consisting of a Prompt Decomposer for extracting spatial metadata, expert detectors for visual grounding of object positions, and a vision-language model with chain-of-thought reasoning to assess complex spatial relations. By focusing on spatial layout evaluation rather than just semantic alignment, this work addresses a critical gap in current reward modeling approaches for generative AI systems.

Anonymous

arXiv 2603.22228v1

cs.RO端到端CV

Make Tracking Easy: Neural Motion Retargeting for Humanoid Whole-body Control

This paper addresses the challenge of transferring human motion data to humanoid robots by proposing Neural Motion Retargeting (NMR), a novel framework that transforms static geometric mapping into a dynamics-aware learned process. The approach uses Clustered-Expert Physics Refinement (CEPR) with VAE-based motion clustering to group heterogeneous movements into latent motifs, significantly reducing computational overhead for reinforcement learning experts that project and repair noisy human motion data. Through Hessian analysis, the authors demonstrate that traditional optimization-based retargeting is inherently non-convex and prone to local optima, leading to physical artifacts. By reformulating the problem as learning data distribution rather than optimizing solutions, the framework achieves smooth, physically plausible whole-body robot control.

NMR Research Team

arXiv 2603.22201v1

cs.CV自动驾驶CV

Feasibility of Augmented Reality-Guided Robotic Ultrasound with Cone-Beam CT Integration for Spine Procedures

This paper presents an optical see-through augmented reality (OST-AR)-guided robotic system for spine procedures, integrating cone-beam CT (CBCT)-derived 3D spine models with live ultrasound for enhanced needle trajectory planning. The system enables in situ visualization of spinal structures, combining global anatomical context from CBCT with local real-time ultrasound feedback. A phantom user study with 16 participants evaluated facet joint injection and lumbar puncture procedures under AR vs. conventional screen visualization conditions, demonstrating the feasibility of AR-guided medical robotics for precise needle placement.

Research Team

arXiv 2603.22174v1

cs.CV大语言模型CV

dynActivation: A Trainable Activation Family for Adaptive Nonlinearity

This paper introduces dynActivation, a per-layer trainable activation function that dynamically interpolates between base nonlinearities and linear paths using lightweight learned scalars. The proposed method is evaluated across vision tasks (CIFAR-10, MNIST) and language modeling tasks, demonstrating significant improvements in training efficiency (up to 54% faster) and performance. On CIFAR-10, dynActivation(Mish) achieves up to 14.02% improvement over static Mish, with 24% reduction in convergence time. In deep network scaling experiments (up to 75 layers), dynActivation maintains robust performance (95.3-99.3% accuracy) while ReLU collapses below 80%, demonstrating that adaptive nonlinearity linearization in deep layers enhances both training stability and final model quality.

dynActivation Authors

arXiv 2603.22154v1

cs.CV自动驾驶CV

OpenEarth-Agent: From Tool Calling to Tool Creation for Open-Environment Earth Observation

This paper presents OpenEarth-Agent, the first tool-creation agent framework designed for open-environment Earth Observation. Unlike traditional tool-calling agents confined to closed environments with pre-defined tools, OpenEarth-Agent employs adaptive workflow planning and dynamic tool creation to generalize across diverse multi-source satellite data and heterogeneous remote sensing tasks. The framework integrates multi-stage tools with cross-domain knowledge bases, enabling robust execution throughout the entire Earth Observation pipeline. This approach represents a significant advancement in autonomous EO systems by moving beyond narrow, predefined toolkits to an open-ended, adaptive framework capable of handling unseen data modalities and novel tasks in land surface monitoring applications.

OpenEarth-Agent Team

arXiv 2603.22148v1

cs.RO端到端CV

ROBOGATE: Adaptive Failure Discovery for Safe Robot Policy Deployment via Two-Stage Boundary-Focused Sampling

This paper presents ROBOGATE, a deployment risk management framework for safe robot policy deployment in industrial settings. The framework combines physics-based simulation with a two-stage adaptive sampling strategy to efficiently discover failure boundaries in high-dimensional operational parameter spaces. Stage 1 uses Latin Hypercube Sampling across an 8-dimensional parameter space, while Stage 2 applies boundary-focused sampling in the 30-70% success rate transition zone. Evaluated using NVIDIA Isaac Sim with Newton physics on Franka Panda and UR5e robots performing pick-and-place tasks across 30,000 experiments, the system employs logistic regression for risk modeling to ensure safe robot manipulation policy deployment.

Robogate Research Team

arXiv 2603.22126v1

cs.CV自动驾驶CV

Future-Interactions-Aware Trajectory Prediction via Braid Theory

This paper presents a novel approach to multi-agent trajectory prediction for autonomous vehicles using braid theory as a fundamental mathematical framework. The method projects future trajectories of interacting agents into braids that precisely describe how trajectories cross over time, enabling exact representation of coordination modes between agents. By fully leveraging the expressivity of the braid representation, the approach conditions predicted trajectories on specific interaction patterns, moving beyond traditional heuristic-based behavior labeling methods. This results in improved joint prediction accuracy while maintaining computational efficiency compared to existing approaches that rely on extensive computational resources.

arXiv 2603.22035v1

cs.CV自动驾驶CV

LRC-WeatherNet: LiDAR, RADAR, and Camera Fusion Network for Real-time Weather-type Classification in Autonomous Driving

Autonomous vehicles encounter significant perception and navigation challenges during adverse weather conditions including rain, fog, and snow, which degrade the performance of LiDAR, RADAR, and RGB camera sensors. This study introduces LRC-WeatherNet, a novel multi-sensor fusion framework that integrates LiDAR, RADAR, and camera data for real-time weather-type classification in autonomous driving. The proposed approach employs both early fusion utilizing a unified Bird's Eye View representation and mid-level gated fusion of modality-specific feature maps, enabling adaptive sensor reliability assessment under varying weather conditions. Evaluated on the extensive MSU-4S dataset covering nine distinct weather types, LRC-WeatherNet demonstrates superior classification performance and computational efficiency, significantly outperforming unimodal baseline approaches.

LRC-WeatherNet Authors

arXiv 2603.21987v1

cs.CV自动驾驶CV

Disengagement Analysis and Field Tests of a Prototypical Open-Source Level 4 Autonomous Driving System

This study presents a comprehensive evaluation of an open-source Level 4 autonomous driving system through extensive field testing across 236 km of mixed traffic scenarios. The research utilizes an Autoware-based software stack and classifies 30 disengagement events using a novel five-level criticality framework, revealing a spatial disengagement rate of 0.127 per kilometer. Analysis demonstrates that perception and planning failures account for 40% and 26.7% of disengagements respectively, with object-tracking losses and operational deadlocks from parked vehicles being primary failure modes. The findings highlight critical challenges in computer vision-based perception and tracking systems for autonomous vehicles operating in complex urban environments.