barquerogerman / belfusion Goto Github PK

Hello, how to train AMASS data sets on GSPS and DiverseSampling, can the scripts related to this part be made public? Thank you very much!

Hi, it seems like the file "segments_test.csv" and "mmapd_GT.csv" files are missing if I just want to perform direct evaluation. Can you please update these files?

Hello,

The file defining the class "PoseTransformerTemporal" seems to be missing. Could you upload it?

Thanks,
Sourav

I want to evaluate your code in another dataset. How can I get this file?

Hello,

I would like to implement Latent Diffusion for pedestrian trajectory forecasting. I am wondering which U-Net class to adapt from. Is it GDUnet_Latent or GDUnet_Latent_vSum?

Awaiting an early response.

Thanks,
Sourav

Thanks!

1

Dear German,

Thanks for your nice work and documentation. When following your installation, I am encountering that running...
python -m data_loader.parsers.amass --gpu

... raises lots of messages regarding files being corrupted.
1957it [01:38, 23.22it/s]WARNING: we skip 'BMLmovi/Subject_5_F_MoSh/shape.npz' because it is corrupted (no framerate)

I have tried to re-download again these tar files from AMASS, but the issue is still there.
Is that behavior expected or did you encounter also that?

Again, congratulations for your nice results

could you please release codes for training baseline models like Dlow? I want to compare them with your method on a new dataset