search

athn-nik / teach

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

Quantitative results correspond to epoch #? #13

Closed chinnusai25 closed 1 year ago

chinnusai25 commented 1 year ago

Can you please confirm the epoch number for which the quantitative results (APE and AVE) are provided in the paper, because it is mentioned as 600 epochs in the paper but the max_epochs in config.yaml of provided pretrained model is 1001

athn-nik commented 1 year ago

I have reported the speed not the performance in Sup.Mat. Use either my checkpoints or if you want to retrain, just use 1000 epochs as a reasonable threshold. My experience was that with batch size 16/32 you would not see big numerical improvements over 600-700 epochs but will improve the Qual. a bit.

chinnusai25 commented 1 year ago

Thanks for the reply, closing the issue