Alert button

Learning Delicate Local Representations for Multi-Person Pose Estimation

Add code
Bookmark button
Alert button
Mar 10, 2020
Yuanhao Cai, Zhicheng Wang, Zhengxiong Luo, Binyi Yin, Angang Du, Haoqian Wang, Xinyu Zhou, Erjin Zhou, Xiangyu Zhang, Jian Sun

Figure 1 for Learning Delicate Local Representations for Multi-Person Pose Estimation
Figure 2 for Learning Delicate Local Representations for Multi-Person Pose Estimation
Figure 3 for Learning Delicate Local Representations for Multi-Person Pose Estimation
Figure 4 for Learning Delicate Local Representations for Multi-Person Pose Estimation

Share this with someone who'll enjoy it:

Twitter IconFacebook IconLinkedin IconWhatsapp IconMessenger Icon

In this paper, we propose a novel method called Residual Steps Network (RSN). RSN aggregates features with the same spatialsize (Intra-level features) efficiently to obtain delicate local representations, which retain rich low-level spatial information and result in pre-cise keypoint localization. In addition, we propose an efficient attention mechanism - Pose Refine Machine (PRM) to further refine the keypointlocations. Our approach won the 1st place of COCO Keypoint Challenge 2019 and achieves state-of-the-art results on both COCO and MPII benchmarks, without using extra training data and pretrained model. Our single model achieves 78.6 on COCO test-dev, 93.0 on MPII test dataset. Ensembled models achieve 79.2 on COCO test-dev, 77.1 on COCO test-challenge dataset. The source code is publicly available for further research at https://github.com/caiyuanhao1998/RSN

View paper onarxiv icon

Share this with someone who'll enjoy it:

Twitter IconFacebook IconLinkedin IconWhatsapp IconMessenger Icon