search

karfly / learnable-triangulation-pytorch

This repository is an official PyTorch implementation of the paper "Learnable Triangulation of Human Pose" (ICCV 2019, oral). Proposed method archives state-of-the-art results in multi-view 3D human pose estimation!
MIT License
1.09k stars 181 forks source link

How to interpret human36m-multiview-labels-GTbboxes.npy? #70

Closed wwang107 closed 4 years ago

wwang107 commented 4 years ago

I want to use my own bounding box for the training. From the way you generate bounding boxes in generate-labels-npy-multiview.py, it seems like you load your own bonding box from a .json file. Do you mind elaborate on the structure of such .json file, so I can save my own bounding box and use the same script? More specifically, does the .json have structure, the correspondence of bounding box -> subject, action, frame_idx, camera ?

Thanks in advance!

shrubb commented 4 years ago

Hi, sorry for a late answer. Does this help?

wwang107 commented 4 years ago

Yes, it does help! thanks a lot! How do you decide on training and testing images?

shrubb commented 4 years ago

You're welcome. See here.

shrubb commented 4 years ago

Oops, closed accidentally

wwang107 commented 4 years ago

Alrighty! thanks for the help. I think I got all I needed for getting your awesome work running.

Again, thanks for the good work and help! you guys are really talented and generous to give help to other researcher or engineer

shrubb commented 4 years ago

Cheers!