Pose-Group / DCPose

This is an official implementation of our CVPR 2021 paper "Deep Dual Consecutive Network for Human Pose Estimation" (https://openaccess.thecvf.com/content/CVPR2021/papers/Liu_Deep_Dual_Consecutive_Network_for_Human_Pose_Estimation_CVPR_2021_paper.pdf)
380 stars 61 forks source link

估计很难实现,在GPU上好像都很慢! #36

Open evinnliu opened 2 years ago

evinnliu commented 2 years ago

估计很难实现,在GPU上好像都很慢!

_Originally posted by @MikuuGithub in https://github.com/Pose-Group/DCPose/issues/24#issuecomment-925449214_Have you solved this problem

chenrxi commented 2 years ago

姿态估计都很慢

Caicaizi-cdy commented 2 years ago

请问大概要多久呢?两块2080/(ㄒoㄒ)/~~感觉数据集超大

chenrxi commented 2 years ago

训练的话要快两天好像,不过我没有复现出结果,才81点多,不过没认真复现。而且因为下巴的坐标是网络直接预测出来的,不像是posewarper利用肩膀的坐标给出下巴的坐标(虽然我也不理解为什么这么做),所以其实直接跟posewarper的结果略有些不公平。如果测试的话提供了检测结果,总体时间比起关节点跟踪的话好多了-_-。 @Caicaizi-cdy

Caicaizi-cdy commented 2 years ago

谢谢回答