2D motion prediction and keypoint format

Question

2D motion prediction and keypoint format

Closed this issue a year ago · 1 comments

Has anybody tried predicting 2D instead of 3D motion? I can see that there's an option to choose depth=2 in the model but I can't find any example of using it.
Also, does the model accept only SMPL keypoint format?

Answer 1 · 2023-11-06T07:37:33.000Z

Hi,
depth=2 represents the depth of the 1D convolutional network in the VQVAE, which does not refer to 2D or 3D.

We have not tried formats other than SMPL, but I think it may be adaptable to other models