Edge Weight Prediction For Category-Agnostic Pose Estimation

TL;DR

EdgeCape introduces a novel approach for category-agnostic pose estimation by predicting edge weights in pose graphs and incorporating structural priors, significantly improving keypoint localization across diverse categories.

Contribution

The paper proposes EdgeCape, a framework that predicts edge weights in pose graphs and integrates Markovian Structural Bias to enhance global spatial dependency modeling.

Findings

Achieves state-of-the-art results on MP-100 benchmark in 1-shot setting.

Outperforms similar methods in 5-shot pose estimation.

Significantly improves keypoint localization accuracy.

Abstract

Category-Agnostic Pose Estimation (CAPE) localizes keypoints across diverse object categories with a single model, using one or a few annotated support images. Recent works have shown that using a pose graph (i.e., treating keypoints as nodes in a graph rather than isolated points) helps handle occlusions and break symmetry. However, these methods assume a static pose graph with equal-weight edges, leading to suboptimal results. We introduce EdgeCape, a novel framework that overcomes these limitations by predicting the graph's edge weights which optimizes localization. To further leverage structural priors, we propose integrating Markovian Structural Bias, which modulates the self-attention interaction between nodes based on the number of hops between them. We show that this improves the model's ability to capture global spatial dependencies. Evaluated on the MP-100 benchmark, which includes 100 categories and over 20K images, EdgeCape achieves state-of-the-art results in the 1-shot setting and leads among similar-sized methods in the 5-shot setting, significantly improving keypoint localization accuracy. Our code is publicly available.