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
155 stars 15 forks source link

H36M, MuCo SMPL GT #27

Open asw91666 opened 1 year ago

asw91666 commented 1 year ago

Hi. thanks for sharing the nice work.

In your code, You use SMPLify-X pseudo GT on 3D datasets(H36M, MuCo), and use Neural Annot pseudo GT on 2D datasets(CrowdPose, MSCOCO). Can I ask why you don't Neural Annot on H36M, MuCo?

hongsukchoi commented 1 year ago

@asw91666

I remember using NeuralAnnot of H36M and MuCo brought marginal gains. I think it's because H36M and MuCo already have accurate 3D joint annotation (not mesh though), since they are MoCap datasets.