vegesm / pose_refinement

MIT License
45 stars 6 forks source link

Error inferring video #7

Open egru5 opened 3 years ago

egru5 commented 3 years ago

pose_refinement.py line:148: poses_init = abs_to_hiprel(pred3d[inds].copy(), joint_set).astype('float32') / 1000

"pred3d" and "inds" are not the same size, this causes an IndexError

I tried to predict a video from the mpi_inf_3dhp set, like in the description stated with: python scripts/predict.py -r video.mp4 output.pkl

So the error is caused, because the Person is not detected on all Frames, this means not all Frames have valid predictions. this causes "pred3d" to be smaller than "inds". Shouldn't the Code be able to handle Frames with no predictions in it? Or am I doing something wrong?