search

kasvii / PMCE

[ICCV 2023] PyTorch Implementation of "Co-Evolution of Pose and Mesh for 3D Human Body Estimation from Video"
https://kasvii.github.io/PMCE
MIT License
145 stars 5 forks source link

Training commands in README #7

Closed zhixuanli closed 7 months ago

zhixuanli commented 7 months ago

Thanks for your awesome codes!

I find that the training commands for the first and second stages are the same, leaving the train_mesh_h36m.sh and train_mesh_3dpw.sh not used. Is this correct?

Here's the current version of the README commands:

Stage 1 (optional): Train the 3D pose estimation stream or you can directly use our pre-traind pose model ./experiment/pretrained/pose_*.pth.tar for Stage 2.

# Human3.6M
bash command/train_pose_h36m.sh

# 3DPW
bash command/train_pose_3dpw.sh

Stage 2: To train the all network for final mesh. Configs of the experiments can be found and edited in ./config folder. Change posenet_path in ./config/train_mesh_*.yml to the path of the pre-trained pose model.

# Human3.6M
bash command/train_pose_h36m.sh

# 3DPW
bash command/train_pose_3dpw.sh

(Second question)

BTW, since there are three datasets used in the paper, could you please release the codes for the "MPI-INF-3DHP" dataset? Thanks for your brilliant work! And looking forward for your replying.

kasvii commented 7 months ago

You are right! The commands in stage 2 should be 'trainmesh*.sh'. Thank you for pointing out this and I have corrected it. We used 'MPII3D' as the short name of 'MPI-INF-3DHP' and you can find the codes here.

zhixuanli commented 7 months ago

Thanks for your fast reply. Now I know how to train the second stage. Thanks!

For the second question, I believe this API code in your link is very helpful. But the ".yml" files for this MPII3D dataset are missing, which makes me not know how to set the hyper-parameters or something else. Could you please add them when you have time?

Have a good night!

kasvii commented 7 months ago

Thanks for your fast reply. Now I know how to train the second stage. Thanks!

For the second question, I believe this API code in your link is very helpful. But the ".yml" files for this MPII3D dataset are missing, which makes me not know how to set the hyper-parameters or something else. Could you please add them when you have time?

Have a good night!

The training datasets for mpii3d are mixed datasets, the same as 3dpw. It can directly use the checkpoints of 3dpw to test ('bash test_mesh_mpii3d.sh'), thus there is no training script for mpii3d.

zhixuanli commented 7 months ago

Thank you for solving all my questions. Hopes everything is fine! :)