Bottom-Up Human Pose Estimation Via Disentangled Keypoint Regression
arXiv:2104.02300
Abstract
In this paper, we are interested in the bottom-up paradigm of estimating human poses from an image. We study the dense keypoint regression framework that is previously inferior to the keypoint detection and grouping framework. Our motivation is that regressing keypoint positions accurately needs to learn representations that focus on the keypoint regions. We present a simple yet effective approach, named disentangled keypoint regression (DEKR). We adopt adaptive convolutions through pixel-wise spatial transformer to activate the pixels in the keypoint regions and accordingly learn representations from them. We use a multi-branch structure for separate regression: each branch learns a representation with dedicated adaptive convolutions and regresses one keypoint. The resulting disentangled representations are able to attend to the keypoint regions, respectively, and thus the keypoint regression is spatially more accurate. We empirically show that the proposed direct regression method outperforms keypoint detection and grouping methods and achieves superior bottom-up pose estimation results on two benchmark datasets, COCO and CrowdPose. The code and models are available at https://github.com/HRNet/DEKR.
Accepted by CVPR2021. arXiv admin note: text overlap with arXiv:2006.15480
References in corpus (7)
- DeepPose: Human Pose Estimation via Deep Neural Networks
- Associative Embedding: End-to-End Learning for Joint Detection and Grouping
- Human pose estimation via Convolutional Part Heatmap Regression
- DirectPose: Direct End-to-End Multi-Person Pose Estimation
- Human Pose Estimation with Spatial Contextual Information
- The Devil is in the Details: Delving into Unbiased Data Processing for Human Pose Estimation
- Peeking into occluded joints: A novel framework for crowd pose estimation