Contrastive-Adaptation-Network-for-Unsupervised-Domain-Adaptation icon indicating copy to clipboard operation
Contrastive-Adaptation-Network-for-Unsupervised-Domain-Adaptation copied to clipboard

Published 20 hours ago • verified publisher icon kgl-prml

Can't reproduce results on VisDA-2017

Open CASlzwde opened this issue 3 years ago • 2 comments

the original paper's result is 87.2, but my average result (the best value of 10 experiment results) is 86.6.

CASlzwde avatar Dec 29 '20 07:12 CASlzwde

Yeah, the same situation also happened to me. I got an average result of 86.5%. Guess the author might have applied some tricks during training process

dingning97 avatar Mar 02 '21 03:03 dingning97

Test at (loop 49, iters: 25399) with mean_accu: 86.8898. I think the result is acceptable.

ygjwd12345 avatar Apr 03 '21 15:04 ygjwd12345