This is a modified version of the following code.
https://akanazawa.github.io/hmr/
https://github.com/akanazawa/hmr
This code enables multi-view training for human mesh recovery.
NOTE: at this point, we didn't see improved results compared to orginal method.
To do training, you should first follow the clone the original github repo and run the demo successfully.
Then follow the train.md in original repo to download all the dataset and models.
Finally replace do_train.sh and ./src/ with my version of code.
Shiyu Mou shiyumou@usc.edu