test with differfent dataset(FG)

[huaitingpp]

   hi! Thank you for your work,I have a question: I train this model with the UTKFace dataset,it's deformation and fuzzy when I test with the FGNetAging dataset,while it's clear when I test with the UTK.
  My purpose is to get a good test results for the  FGNetAging dataset.And,I process FG dataset with alignment before.
  So,I try some methods from different train dataset and epochs:

(1)train dataset:UTK+FG,a total of 23708+1002=24710; epochs:50; test result:as follows
you can see:the test of FG is deformation and fuzzy.while ,the test of FG is clear.
input_FG.png:

test_as_female_FG.png:

test_as_male_FG.png:

(2)train dataset:FG,a total of 1002; epochs:200; test result:as follows
input_FG.png:

test_as_female_FG.png:

test_as_male_FG.png:

I try to increase the epoch,but the result also is not good.Maybe I should adjust other parameters?Can you tell me some advice.Look forward to your reply.thank you!

[Romeo-CC]

have you solved this problem?
and I also want to train by FG ,can you give me your labled data?
my email is na_liu1994@126.com
thank you very much !

[liyin2015]

how did you train the network? My result from the original network the generated male and female images are all the same

[liushanbing]

Hi@huaitingpp
I also want to get a good test results on the other dataset,which has 2000 images of 64 * 64 .I have trained on UTKFace with the author‘s code and got good results . But the results are not good on my datasets.I think mabye the data is insufficient ,so I want to merging UTKFace with FG training ,can you give me your labled data?my email is liushanbing@bit.edu.cn.thank you very much !