shunlinlu
shunlinlu
And also, it'll be better if you can release the zero-shot classification influence script for RSNA datasets.
I met the same problem. Any solution?
Hi, @sohananisetty This initialization actually does not matter. You can test using your initialization. And with my visualization, it seems good. I think your initialization is also fine.  Shunlin
Hi, @andrewnc We actually do not use rotation in our representation from our experiments in the appendix. And we followed Humanml3D format to visualize the motion using position. If you...
Hi, @maemoKKK Did you find a way to solve it? Shunlin
I trained exactly using this command: `python -m src.train.train --clip_text_losses cosine --clip_image_losses cosine --pose_rep rot6d \ --lambda_vel 100 --lambda_rc 100 --lambda_rcxyz 100 \ --jointstype vertices --batch_size 20 --num_frames 60 --num_layers...
And I test the model you provided, it works fine.
> By rotation part, does it mean Joint Rotation representation (rot_data) in Line 308 in motion representation.py? Isn't it still in the final representation? > > line 308: rot_data =...
Did you solve this problem? @lxycopper