ViTPose-Pytorch icon indicating copy to clipboard operation
ViTPose-Pytorch copied to clipboard

WholeBody models

Open oscarfossey opened this issue 2 years ago • 7 comments

Hi, thank you for your work! Do you know if the current implementation would work with the large wholdebody_model from Vit_pose?

oscarfossey avatar Jun 06 '23 15:06 oscarfossey

Hello, I am not sure I can find pretrained weights to test it. Nonetheless, I will make some changes to the repo. The package version works with only vitbpose b right now.

gpastal24 avatar Jun 06 '23 17:06 gpastal24

Alright it's clear. If you need them the large wholebody weights here : large_wholebody The weight's link are at the bottom of the readme of the official repo. In the "WholeBody" part. For small, base, large and huge models. It says that you have to split them with their script it works well. But using mmpose implies to have an mmcv dependencie :no_good: .

oscarfossey avatar Jun 07 '23 08:06 oscarfossey

Hi ok I thought vitpose+ was a different architecture. I will look into this the coming weekend and let you know . I ll probably need to find/write some visualization functions for this case as well.

gpastal24 avatar Jun 08 '23 20:06 gpastal24

Hi By chance any update on this ?

keniel123 avatar Mar 05 '25 23:03 keniel123