Heatmap-based methods dominate in the field of human pose estimation by modelling the output distribution through likelihood heatmaps . Residual Log-likelihood Estimation(RLE) is effective, efficient and flexible . Compared to the conventional regression paradigm, regression with RLE bring 12.4 mAPimprovement on MSCOCO without any test-time overhead . For the firsttime, especially on multi-person pose estimation, our regression method is superior to the heatmap-style methods for the first time . Our code is available athttps://://://github.com/Jeff-sjtu/res-loglikelihood-regression-regressment-rearchenginging-researche.com. We show its potential in varioushuman pose estimation tasks with comprehensive experiments. The proposedmethod is effective and flexible. We showed its potential

Author(s) : Jiefeng Li, Siyuan Bian, Ailing Zeng, Can Wang, Bo Pang, Wentao Liu, Cewu Lu

Links : PDF - Abstract

Code :
Coursera

Keywords : estimation - regression - pose - likelihood - rle -

Leave a Reply

Your email address will not be published. Required fields are marked *