Releases: m-decoster/ChaLearn-2021-LAP
Pre-trained models
These are the models for my final predictions on the leaderboards.
There is one model for the RGB track and one for the RGB-D track.
They can be extracted using tar -jxvf CP_FILE.tar.bz2
, resulting in a .ckpt
file, which can be loaded by PyTorch Lightning.
To use these models for predictions, see README.md
.
Keypoints
This release contains data files for the keypoint and pose flow features.
The archives have been split because of file size constraints. They can be reassembled using, e.g., the following command: cat kp.tar.bz2.part* > kp.tar.bz2
after which you can extract the files using tar -jxvf kp.tar.bz2
. Alternatively you can directly pipe to tar: cat kp.tar.bz2.part* | tar -jxv
.
You should do this for both kp.tar.bz2
and kpflow2.tar.bz2
.
For more information on where to place the resulting files, see README.md
.