search

terarachang / ACCV_TinyGAN

BigGAN; Knowledge Distillation; Black-Box; Fast Training; 16x compression
75 stars 9 forks source link

Results on whole ImageNet dataset #4

Closed vominhduc closed 3 years ago

vominhduc commented 3 years ago

Hi Chang,

Thank you for your interesting work. I have tried your method on the whole ImageNet dataset (1000 classes) but the results are weird. I am wondering if you are successful in dealing with 1000 classes?

Best, Duc.

terarachang commented 3 years ago

Hi,

Unfortunately, our results on all 1000 classes are also miserable. Please see our paper (page 21) for more discussion if you are interested: https://arxiv.org/pdf/2009.13829.pdf

Thanks, Ting-Yun