search

jewelltaylor / OLED

6 stars 1 forks source link

poor result in CIFAR-10 #1

Open tjl1 opened 2 years ago

tjl1 commented 2 years ago

Hi ! I tried to use the data set and weight file you provided for the test, and I could get the effect you said in the paper on the MNIST data set, but the effect on the CIFAR data set was particularly poor. The following are the results I obtained. Is there anything wrong with what I did?

input-img: image image

mask-img: image image

decoder_mask_output: image image

decoder_normal_output: image image

I get the images with the code in test_step(): image

looking forward to your reply !

jewelltaylor commented 2 years ago

Will take a look into this! This is definitely not the expected behaviour