search

Nicholasli1995 / EvoSkeleton

Official project website for the CVPR 2020 paper (Oral Presentation) "Cascaded deep monocular 3D human pose estimation wth evolutionary training data"
https://arxiv.org/abs/2006.07778
MIT License
333 stars 44 forks source link

How to make 2d pose estimation output .npy? #29

Closed limeverry closed 3 years ago

limeverry commented 3 years ago

Hello. I want to run in the wild data using inference. but I don't know how to make 2d pose estimation output (.npy file). According to what I understand, For making npy file, I should use h36m2Dpose.py or annotate_2D.p, right?

Can you explain how?

thank you.

Nicholasli1995 commented 3 years ago

Hello. I want to run in the wild data using inference. but I don't know how to make 2d pose estimation output (.npy file). According to what I understand, For making npy file, I should use h36m2Dpose.py or annotate_2D.p, right?

Can you explain how?

thank you.

Hi,

  1. If you want to use h36m2Dpose.py, you can refer to https://github.com/Nicholasli1995/EvoSkeleton/blob/master/docs/2DHPE.md. The 2D pose outputs are visualized in that example. You can save the 2D pose predictions with np.save().
  2. If you want to use annotate_2D.py, you can refer to https://github.com/Nicholasli1995/EvoSkeleton/blob/master/docs/ANNOTATOR.md. Pressing Z will save your annotated 2D key-points.
limeverry commented 3 years ago

Thank you for your help!

In my case, I want to use h36m2dpose.py . I think It is necessary to add 'np.save('filename.npy', pred)' on 120 line in h36m2Dpose.py. right?

In this case I have one concern. if I make .npy file as above, Does this result match the format of 3d pose estimation? In the case of 3d PE, I know that it has the same key as 'p2d'.

Nicholasli1995 commented 3 years ago

Thank you for your help!

In my case, I want to use h36m2dpose.py . I think It is necessary to add 'np.save('filename.npy', pred)' on 120 line in h36m2Dpose.py. right?

In this case I have one concern. if I make .npy file as above, Does this result match the format of 3d pose estimation? In the case of 3d PE, I know that it has the same key as 'p2d'.

Hi, the inference example in https://github.com/Nicholasli1995/EvoSkeleton/blob/master/docs/TRAINING.md uses 16 joints while h36m2Dpose.py outputs 17 joints. You need to remove the nose joint for inference. You can find the order of the input joints in the inference example script.

Nicholasli1995 commented 3 years ago

Closing inactive issues. Re-open if you have further questions.