search

hongsukchoi / 3DCrowdNet_RELEASE

Official Pytorch implementation of "Learning to Estimate Robust 3D Human Mesh from In-the-Wild Crowded Scenes", CVPR 2022
MIT License
156 stars 14 forks source link

H36M evaluation #29

Open mimiliaogo opened 1 year ago

mimiliaogo commented 1 year ago

Hi, Have you ever evaluated your method on Human3.6M testing set? Because I am curious about whether your method still works well in simple environments and is comparable to other methods.

hongsukchoi commented 1 year ago

@mimiliaogo

Yes, I remember it worked well. I guess there's no reason for 3DCrowdNet to work worse in simple environments.

mimiliaogo commented 1 year ago

@hongsukchoi
Thank you for getting back to me. But I used your model weight given in here with version 04-06 snapshot 10, and evaluated it on Human3.6M but got a poor result compared to previous methods: MPJPE from mesh: 69.22 mm PA MPJPE from mesh: 44.47 mm

Do you have any idea about this result?

Thank you so much.

hongsukchoi commented 1 year ago

@mimiliaogo

The most likely reason is using wrong 2D pose input. Which 2D pose input did you use?

MilkLoo commented 7 months ago

@mimiliaogo Hello, I would like to ask you how to write the code for the test and evaluation part of Human3.6M? Can you share? Thank you very much! 2395611610@qq.com