Closed Oocean35 closed 9 months ago
Hi, I think for the video dataset, you need to first estimate the camera and SMPL parameters. Maybe one possible solution is to use CLIFF (CLIFF: Carrying Location Information in Full Frames into Human Pose and Shape Estimation). Then you can follow our RenderPeople dataloader to train the model.
Hi @skhu101 , Thanks a lot for sharing this very insightful work ! It would be really great if you could provide the code that one can use to prepare one's own video dataset to use for training ! This would be incredibly helpful ! Thanks a lot !