search

mkocabas / VIBE

Official implementation of CVPR2020 paper "VIBE: Video Inference for Human Body Pose and Shape Estimation"
https://arxiv.org/abs/1912.05656
Other
2.85k stars 551 forks source link

[BUG] Fbx conversion script only giving out one person #289

Open RohaanA opened 11 months ago

RohaanA commented 11 months ago

Thanks for your interest in our research!

If you have problems running our code, please include;

  1. your operating system and the version: Windows 11 x64 bit
  2. your python version, Python 3.7
  3. your pytorch version, 1.4.0
  4. the stack trace of the error that you see, No error

Specifically, if you have an issue with pyrender or OpenGL setup & installation, please refer to pyrender docs or github issues.

Hi, I ran the FbX conversion script on the pkl file the model outputted. However, I can only see one guy in the FBx file, while the model had tracked 2-3 people in the video. Is it a limitation of 1 person?

elia-img commented 5 months ago

@RohaanA I think you are supposed to use the --person_id flag specifying the <tracklet id from VIBE output> maybe ? Although my question is where do I get the tracklet IDs from... I can't seem to be able to open the .pkl file

RohaanA commented 5 months ago

@RohaanA I think you are supposed to use the --person_id flag specifying the <tracklet id from VIBE output> maybe ? Although my question is where do I get the tracklet IDs from... I can't seem to be able to open the .pkl file

Yup, that is what I did as well. For the IDs you can simply try starting from 0. The tracklet IDs start from 0 and go up depending on how many people are tracked in the video

elia-img commented 5 months ago

@RohaanA I think you are supposed to use the --person_id flag specifying the <tracklet id from VIBE output> maybe ? Although my question is where do I get the tracklet IDs from... I can't seem to be able to open the .pkl file

Yup, that is what I did as well. For the IDs you can simply try starting from 0. The tracklet IDs start from 0 and go up depending on how many people are tracked in the video

Oh that makes a lot of sense. Thanks for the help :)