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

json #4

Closed HuiliCui closed 2 years ago

HuiliCui commented 2 years ago

Thanks for your share！ If I want to train with my dataset, how can I get the json file about 2D keypoints?

hongsukchoi commented 2 years ago

During training 3DCrowdNet uses GT 2D keypoints. Does your dataset has 2D keypoints?

HuiliCui commented 2 years ago

There are 2D joint points generated by alphapose

---Original--- From: "Hongsuk @.> Date: Fri, May 6, 2022 18:02 PM To: @.>; Cc: @.**@.>; Subject: Re: [hongsukchoi/3DCrowdNet_RELEASE] json (Issue #4)

During training 3DCrowdNet uses GT 2D keypoints. Does your dataset has 2D keypoints?

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: @.***>

hongsukchoi commented 2 years ago

You can use them instead of GT 2D keypoints durining training. But make sure to eliminate PoseFix code which is adding realistic error to the GT 2D keypoints.

HuiliCui commented 2 years ago

Thanks for your reply. I have one more question. Can I use the 2D joint points generated by alphapose during the test?

---Original--- From: "Hongsuk @.> Date: Fri, May 6, 2022 18:10 PM To: @.>; Cc: @.**@.>; Subject: Re: [hongsukchoi/3DCrowdNet_RELEASE] json (Issue #4)

You can use them instead of GT 2D keypoints durining training. But make sure to eliminate PoseFix code which is adding realistic error to the GT 2D keypoints.

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: @.***>

hongsukchoi commented 2 years ago

Yes you can. Check this joint order before training and testing. https://github.com/hongsukchoi/3DCrowdNet_RELEASE/blob/c4d32c0dbb2675b022f9bc2848a39242d1142244/common/utils/smpl.py#L36

Not every joint must exist. Just set missing joints's joint_valid to be zero.

HuiliCui commented 2 years ago

OK, thank you again for your timely reply！

---Original--- From: "Hongsuk @.> Date: Fri, May 6, 2022 18:23 PM To: @.>; Cc: @.**@.>; Subject: Re: [hongsukchoi/3DCrowdNet_RELEASE] json (Issue #4)

Yes you can. Check this joint order before training and testing. https://github.com/hongsukchoi/3DCrowdNet_RELEASE/blob/c4d32c0dbb2675b022f9bc2848a39242d1142244/common/utils/smpl.py#L36

Not every joint must exist. Just set missing joints's joint_valid to be zero.

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: @.***>

hongsukchoi commented 2 years ago

Sure:)