search

Walter0807 / MotionBERT

[ICCV 2023] PyTorch Implementation of "MotionBERT: A Unified Perspective on Learning Human Motion Representations"
Apache License 2.0
1.01k stars 123 forks source link

Finetune data format #49

Closed XALEX-123 closed 1 year ago

XALEX-123 commented 1 year ago

hello @Walter0807 , I wanna finetune pose3d task for my own dataset, what data format should I prepare, like what's inside .pkl file. Now I have a 2D skeleton video and the json file from Alphapose, what should I do next. Sorry for keeping bothering you.

Walter0807 commented 1 year ago

Hi, do you want to finetune on a 2D dataset or inference on your own videos with 2D pose estimation results?

XALEX-123 commented 1 year ago

I want to finetune the process of lifting 2D to 3D I have Azure Kinect so I can create the 3D ground truth on my own. However, I don't know those number in .pkl mean, can you share some information to me? That's will be a huge help, thanks!!

Walter0807 commented 1 year ago

You can check the data processing part in LCN. Basically, you need to align 3D GT and 2D detections in the pixel coordinates so that you can finetune the 2D->3D mapping.