search

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

Pose Temporal Merger #11

Closed boza-wd closed 3 years ago

boza-wd commented 3 years ago

Hi folks,

thanks for your amazing work.

I have a question regarding the results I have obtained using DCPose.

Below you can see three consecutive images (from a longer sequence). It seems that the temporal component does not correct the false association (left/right knee) which it could have if it used the information from the previous and the next frame. Am I using it in a wrong way, or you expected this kind of behaviour.

00000040 00000041 00000042

Cheers!

chenhaomingbob commented 3 years ago

Hi folks,

thanks for your amazing work.

I have a question regarding the results I have obtained using DCPose.

Below you can see three consecutive images (from a longer sequence). It seems that the temporal component does not correct the false association (left/right knee) which it could have if it used the information from the previous and the next frame. Am I using it in a wrong way, or you expected this kind of behaviour.

00000040 00000041 00000042

Cheers!

Hi boze-wd: Sorry, our DCPose occasionally happens the error that discontinuities of human poses, but it works well in most cases.