using Megaage-Asian train mae is 3.8,is this normal？

Question

using Megaage-Asian train mae is 3.8,is this normal？

zys1994 opened this issue 5 years ago · 4 comments

zys1994 commented 5 years ago

i train ssrnet with Megaage-Asian and get the mae is 3.8,is this normal？ what's your result about Megaage-Asian. Thanks

Answer 1 · 2019-08-16T10:09:38.000Z

Hello. We didn't test MAE on Megaage-Asian. It was CA on the other repository.

Answer 2 · 2019-12-06T07:05:17.000Z

Hi @zys1994 I trained with Megaage-Asian and get mae is 3.91 which is lower than you. Have you ever change the original code somewhere? Thanks

Answer 3 · 2019-12-06T07:32:15.000Z

We didn't test MAE on Megaage-Asian. It was CA on the other repository.
BTW, the batchsize and the version of the tensorflow and keras you use might affect the results too.

Answer 4 · 2019-12-06T07:41:01.000Z

Thanks for your reply, @shamangary . Maybe you are right because I also get the CA result on MegaAge-Asian, which CA3=0.536, CA5=0.731. It is sight lower than your result: CA3=0.549, CA5=0.741. I didn't change the code except some modification caused by keras version.