cs.CV（2025-08-28）

📊 共 34 篇论文 | 🔗 14 篇有代码

🎯 兴趣领域导航

支柱九：具身大模型 (Embodied Foundation Models) (13 🔗6) 支柱二：RL算法与架构 (RL & Architecture) (6) 支柱三：空间感知与语义 (Perception & Semantics) (5 🔗1) 支柱七：动作重定向 (Motion Retargeting) (4 🔗4) 支柱一：机器人控制 (Robot Control) (4 🔗2) 支柱四：生成式动作 (Generative Motion) (2 🔗1)

🔬 支柱九：具身大模型 (Embodied Foundation Models) (13 篇)

#题目一句话要点标签🔗⭐
1 CogVLA: Cognition-Aligned Vision-Language-Action Model via Instruction-Driven Routing & Sparsification CogVLA：通过指令驱动的路由和稀疏化实现认知对齐的视觉-语言-动作模型 vision-language-action VLA multimodal ✅
2 Dino U-Net: Exploiting High-Fidelity Dense Features from Foundation Models for Medical Image Segmentation Dino U-Net：利用DINOv3高保真密集特征提升医学图像分割精度 foundation model ✅
3 PathMR: Multimodal Visual Reasoning for Interpretable Pathology Diagnosis 提出PathMR：用于可解释病理诊断的多模态视觉推理框架 multimodal ✅
4 Adapting Foundation Model for Dental Caries Detection with Dual-View Co-Training 提出DVCTNet，利用双视角协同训练提升牙齿龋齿检测精度 foundation model ✅
5 Graph-Based Uncertainty Modeling and Multimodal Fusion for Salient Object Detection 提出基于图的不确定性建模与多模态融合的显著性目标检测网络，提升复杂场景下的检测精度。 multimodal ✅
6 MedFoundationHub: A Lightweight and Secure Toolkit for Deploying Medical Vision Language Foundation Models MedFoundationHub：轻量安全医学视觉语言模型部署工具包，解决PHI暴露风险。 foundation model
7 Veritas: Generalizable Deepfake Detection via Pattern-Aware Reasoning 提出Veritas，通过模式感知推理实现深度伪造检测的泛化性，并使用HydraFake数据集进行评估。 large language model chain-of-thought
8 R-4B: Incentivizing General-Purpose Auto-Thinking Capability in MLLMs via Bi-Mode Annealing and Reinforce Learning R-4B：通过双模退火和强化学习，激励MLLM的通用自动思考能力 large language model multimodal
9 GENNAV: Polygon Mask Generation for Generalized Referring Navigable Regions GENNAV：用于广义指代可导航区域的多边形掩码生成 zero-shot transfer
10 Generalizable Object Re-Identification via Visual In-Context Prompting 提出基于视觉上下文提示的通用物体重识别方法，无需特定类别训练。 foundation model ✅
11 MMG-Vid: Maximizing Marginal Gains at Segment-level and Token-level for Efficient Video LLMs MMG-Vid：通过分段和Token级最大化边际收益，提升视频LLM效率 large language model
12 Looking Beyond the Obvious: A Survey on Abstract Concept Recognition for Video Understanding 综述：视频抽象概念识别，利用基础模型促进视频理解 foundation model
13 Improving Alignment in LVLMs with Debiased Self-Judgment 提出基于去偏自我判断的LVLM对齐方法，提升视觉语言模型的安全性和准确性。 large language model

🔬 支柱二：RL算法与架构 (RL & Architecture) (6 篇)

#题目一句话要点标签🔗⭐
14 "Humor, Art, or Misinformation?": A Multimodal Dataset for Intent-Aware Synthetic Image Detection 提出S-HArM数据集，用于意图感知的合成图像检测，解决现有方法忽略图像生成意图的问题。 contrastive learning HuMoR multimodal
15 HiddenObject: Modality-Agnostic Fusion for Multimodal Hidden Object Detection 提出HiddenObject，利用Mamba融合RGB、深度和热成像数据，提升隐藏物体检测性能。 Mamba multimodal
16 OneReward: Unified Mask-Guided Image Generation via Multi-Task Human Preference Learning OneReward：通过多任务人类偏好学习统一的掩码引导图像生成 reinforcement learning preference learning
17 PHD: Personalized 3D Human Body Fitting with Point Diffusion PHD：利用点扩散的个性化3D人体姿态拟合，提升视频姿态估计精度。 distillation human mesh recovery HMR
18 Contrastive Learning through Auxiliary Branch for Video Object Detection 提出CLAB方法，通过对比学习辅助分支提升视频目标检测的鲁棒性。 contrastive learning
19 Enhancing Mamba Decoder with Bidirectional Interaction in Multi-Task Dense Prediction 提出双向交互Mamba（BIM）模型，提升多任务密集预测的性能。 Mamba

🔬 支柱三：空间感知与语义 (Perception & Semantics) (5 篇)

#题目一句话要点标签🔗⭐
20 ${C}^{3}$-GS: Learning Context-aware, Cross-dimension, Cross-scale Feature for Generalizable Gaussian Splatting 提出C³-GS以解决高质量视图合成中的特征编码不足问题 gaussian splatting splatting ✅
21 Adam SLAM - the last mile of camera calibration with 3DGS Adam SLAM：利用3DGS微调相机标定，提升新视角合成质量 3DGS NeRF
22 Realistic and Controllable 3D Gaussian-Guided Object Editing for Driving Video Generation 提出G^2Editor，用于自动驾驶视频中逼真且可控的3D高斯引导物体编辑 3D gaussian splatting gaussian splatting splatting
23 Enhancing Pseudo-Boxes via Data-Level LiDAR-Camera Fusion for Unsupervised 3D Object Detection 提出数据级LiDAR-Camera融合方法，用于无监督3D目标检测，显著提升伪标签质量。 depth estimation foundation model
24 AvatarBack: Back-Head Generation for Complete 3D Avatars from Front-View Images AvatarBack：提出一种从正面图像生成完整3D头像背面头部的新框架 gaussian splatting splatting

🔬 支柱七：动作重定向 (Motion Retargeting) (4 篇)

#题目一句话要点标签🔗⭐
25 SeqVLM: Proposal-Guided Multi-View Sequences Reasoning via VLM for Zero-Shot 3D Visual Grounding SeqVLM通过多视角序列推理和VLM，实现零样本3D视觉定位 spatial relationship visual grounding ✅
26 COMETH: Convex Optimization for Multiview Estimation and Tracking of Humans COMETH：基于凸优化的多人多视角人体姿态估计与跟踪 human motion motion tracking ✅
27 SYNBUILD-3D: A large, multi-modal, and semantically rich synthetic dataset of 3D building models at Level of Detail 4 SYNBUILD-3D：一个大规模、多模态、语义丰富的LoD4级别3D建筑模型合成数据集。 geometric consistency ✅
28 FW-GAN: Frequency-Driven Handwriting Synthesis with Wave-Modulated MLP Generator FW-GAN：基于频率驱动和波调制MLP生成器的手写体合成 spatial relationship ✅

🔬 支柱一：机器人控制 (Robot Control) (4 篇)

#题目一句话要点标签🔗⭐
29 DrivingGaussian++: Towards Realistic Reconstruction and Editable Simulation for Surrounding Dynamic Driving Scenes DrivingGaussian++：面向自动驾驶场景的真实重建与可编辑模拟 manipulation scene reconstruction large language model ✅
30 Webly-Supervised Image Manipulation Localization via Category-Aware Auto-Annotation 提出基于类别感知的自动标注Webly监督图像篡改定位方法，缓解数据稀缺问题。 manipulation ✅
31 Enhancing Corpus Callosum Segmentation in Fetal MRI via Pathology-Informed Domain Randomization 提出病理信息驱动的域随机化方法，提升胎儿MRI中胼胝体分割精度，尤其针对胼胝体发育不全。 domain randomization
32 Towards Mechanistic Defenses Against Typographic Attacks in CLIP 针对CLIP中印刷攻击，提出一种基于选择性消融的防御机制。 manipulation

🔬 支柱四：生成式动作 (Generative Motion) (2 篇)

#题目一句话要点标签🔗⭐
33 Embracing Aleatoric Uncertainty: Generating Diverse 3D Human Motion Diverse-T2M：通过引入不确定性生成多样化3D人体运动 text-to-motion motion generation human motion
34 EmoCAST: Emotional Talking Portrait via Emotive Text Description EmoCAST：提出一种基于扩散模型的文本驱动情感化说话人像生成框架 motion synthesis ✅

⬅️ 返回 cs.CV 首页 · 🏠 返回主页