search

hongsukchoi / Pose2Mesh_RELEASE

Official Pytorch implementation of "Pose2Mesh: Graph Convolutional Network for 3D Human Pose and Mesh Recovery from a 2D Human Pose", ECCV 2020
MIT License
678 stars 69 forks source link

Joints order of the parsed data #48

Open aali26 opened 2 years ago

aali26 commented 2 years ago

Hello Choi,

Thanks for the wonderful contribution. I was wondering what is joints order of the parsed data (17, 2). You mentioned in another issue you used (Gyeongsik Moon) work for parsing the data but I'm not sure which project exactly.

Thanks in advanced for your response

hongsukchoi commented 2 years ago

Hi,

It’s inthe ‘./data/*/dataset.py’

aali26 commented 2 years ago

Thank you Choi for your comment. What I want to do is to experiment different 2d detector which requires me to understand the joints order your code accepts. So I can mimic the same input. I know the file you mentioned the joints order that you use.

One more question let’s say I want to use openpose detector which has more joints than what you have, do you think it is good idea to just do static mapping between the joints order of openpose and yours and then retrain the model or you think this is not practical?

hongsukchoi commented 2 years ago

One more question let’s say I want to use openpose detector which has more joints than what you have, do you think it is good idea to just do static mapping between the joints order of openpose and yours and then retrain the model or you think this is not practical?

It will work fine. But in the test time, there should no missing joints in the input. See my 3DCrowdNet repo for advanced estimation