DCLGAN icon indicating copy to clipboard operation
DCLGAN copied to clipboard

mode collapse

Open hd201708010401 opened this issue 2 years ago • 4 comments

I have some problems with my training facdes dataset and maps dataset 1 1

The translation result on facdes is always the same, and there will be no change 1_A 1_A

The translation results on the maps dataset are shown in the figure below.

hd201708010401 avatar Jul 13 '22 02:07 hd201708010401

Hello, Yes, DCLGAN suffers from mode collapses in such datasets (see section 5.3 of the paper for a discussion).

CycleGAN/SimDCL might performs better.

JunlinHan avatar Jul 13 '22 04:07 JunlinHan

@JunlinHan Can you provide the pre-training model of SimDCL trained facades? My training model still has the phenomenon of model collapse, which cannot reproduce the effect in the article.

hd201708010401 avatar Sep 24 '22 01:09 hd201708010401

@hd201708010401 Hello, Oh I do not have a copy of this pre-trained model.

Did you use the default training setting? If possible, could you share your training log? Be careful with the number of GPU and batchsize.

JunlinHan avatar Sep 24 '22 07:09 JunlinHan

邮件已经收到,谢谢。I have received your email. Thank you for your letter.

hd201708010401 avatar Sep 24 '22 07:09 hd201708010401