Annotate video with and extract keypoints instead of mesh
Closed this issue · 2 comments
adelaneh commented
Hi,
Is it possible to export the keypoints from a video in `track.py` instead of mesh?
Thanks,
-Adel
geopavlakos commented
You can readout the 3D keypoints directly from our output.
If you are using the tracking functionality track.py
, these are stored as part of the output pkl file, e.g., your_output_pkl['your_frame_name']['3d_joints']
.
If you are using the single frame functionality demo.py
, you can read them from the out
variable here, e.g., out['pred_keypoints_3d']
.
adelaneh commented
Thanks a lot, @geopavlakos