Yu Zhu, Dan Zeng, Shuiwang Li et al. (6 total)
2025-11-17
ArXiv Vol. abs/2511.13102
10.48550/arxiv.2511.13102
摘要
Recent research in Category-Agnostic Pose Estimation (CAPE) has adopted fixed textual keypoint description as semantic prior for two-stage pose matching frameworks. While this paradigm enhances robustness and flexibility by disentangling the dependency of support images, our critical analysis reveals two inherent limitations of static joint embedding: (1) polysemy-induced cross-category ambiguity during the matching process(e.g., the concept "leg" exhibiting divergent visual manifestations acros...
该研究旨在探索一种新的姿态提议任务，即模型无需任何支持，直接为类别无关的对象提议关键点和链接，同时保证结构适应性和语义一致性，从而解决现有方法对标注和分类的依赖问题。
本文提出了一种基于协同匹配监督的姿态细化框架，该框架联合学习来自基础类别的可迁移关键点和链接，从而为类别无关的对象提议姿态。该框架包含多个细化层，逐步获得精确的姿态。
在大型多类别姿态数据集MP-100上的大量实验和深入分析表明，该方法是有效的。与现有方法相比，该方法在点 mAP 和链接 mAP 方面均取得了显著提升，表明了姿态细化框架和协同匹配监督的有效性。