ZZUTK / Face-Aging-CAAE

Age Progression/Regression by Conditional Adversarial Autoencoder
636 stars 233 forks source link

I get a blank result ,could you please to help me? #30

Open yl-zhu opened 6 years ago

yl-zhu commented 6 years ago

@ZZUTK Hi~ I just try to run this code with several version tensorflow, but none of them are able to get the expected result.

The problem as follow:

  1. First,I use tensorflow1.2 to run the code, then I change code: (1) tf.concat (ops.py) (2) tf.sigmoid_~~~(Face-Aging.py) (3) tf.train.AdamOptimizer --> tf.train.GradientDescentOptimizer (Face-Aging.py)

This code can work successfully. But the result it produce are all blank(UTKFace dataset), just like the image below. I try to change back to AdamOptimizer,but it can not even run. So I do not know how to solve this problem,could you please help me?

28

  1. Then I install thensorflow 0.5, and just run the code, but I find that this project have used some def non-existent in the original 0.5 version tensorflow, such as tf.nn.conv2d_transpose(ops.py). It is hard for me to complement it by myself.
  2. Later,I surf github to find your answer #20 ‘Update to tensorflow 1.4’,because it is hard to download version 1.4 so far, so I try it on tensorflow 1.2(does not exist parameter reuse choise AUTO_REUSE) and tensorflow 1.7(caused some dimension fault), in a word, it work failed.

Above all, if there any method to make it to product the right result rather than blank image result, or how to upgrade the code ?

praveenkumarchandaliya commented 6 years ago

I am also getting the blank image in sample and test folder 01

praveenkumarchandaliya commented 6 years ago

Update Version solved this problem