search

athn-nik / teach

Official PyTorch implementation of the paper "TEACH: Temporal Action Compositions for 3D Humans"
https://teach.is.tue.mpg.de
Other
384 stars 40 forks source link

problem in amass_splits_babel.py #15

Closed zhuangzhuang000 closed 1 year ago

zhuangzhuang000 commented 1 year ago

I ran the amass_splits_babel.py , and get the result "Percentage not found: 879/1412", is that correct?

athn-nik commented 1 year ago

can you give me more details about your bug? Is the process amass correctly finished? What are its results? After that what is the exact command you run? and what the error log?

zhuangzhuang000 commented 1 year ago

can you give me more details about your bug? Is the process amass correctly finished? What are its results? After that what is the exact command you run? and what the error log?

I think I have found out why, when running process_amass.py, it be killed in the middle of the data processing, I continue to run, but the result amass.pth.tar file will be overwritten, so it does not contain all the data. I'm trying to get the amass.pth.tar file again.

athn-nik commented 1 year ago

Yes, better delete that and start fresh. In that case, you should have the correct file. Let me know how it goes. Worst case, I will try to find a way to upload the processed data directly.

athn-nik commented 1 year ago

I hope this fixes the problem. If so, ping here so I can close this.

zhuangzhuang000 commented 1 year ago

I hope this fixes the problem. If so, ping here so I can close this.

My computer crashed so I was delayed for a few days, the code was always killed when I was running process_amass.py, I found that the reason was not enough memory (amass data is too large), after several attempts, I finally got the complete amass.pth.tar data, re-run amass_splits_babel.py and the result is shown in the following figure. I think this should be correct.

image

athn-nik commented 1 year ago

Yes :) the script seems to run smoothly. I just looked at it. The percentage not found is just a statistic for the extra amass that are not included/labelled in BABEL. This will be probably done in the future