← all repositories

caiyuanhao1998/RSN

PyTorch implementation of Residual Steps Network (RSN) for multi-person human pose estimation achieving 1st place in COCO 2019 Keypoint Challenge.

504 stars Python Computer Vision
RSN
Velocity · 7d
+0.2
★ / day
Trend
steady
star history

This repository implements RSN, a deep learning approach for detecting human body keypoints in multi-person images. The method aggregates features at the same spatial resolution to obtain delicate local representations that retain spatial information for precise keypoint localization. It includes a Pose Refine Machine attention mechanism to further refine keypoint predictions. The model achieved top performance on COCO dataset benchmarks.

heatdrop uses Google Analytics to see which pages get read — nothing else. Your call. How we handle data.