Does it only test the input size (1280, 720) ?

dongheehand / DeblurGAN-tf

Tensorflow implementation of DeblurGAN(Blind Motion Deblurring Using Conditional Adversarial Networks)

80 stars 24 forks source link

Does it only test the input size (1280, 720) ? #6

Open chenyuqing opened 5 years ago

chenyuqing commented 5 years ago

When I test a small image, it occure this error.

InvalidArgumentError (see above for traceback): Incompatible shapes: [1,48,44,3] vs. [1,47,41,3] [[node generator/add_32 (defined at /home/ares2/workspace/DeblurGAN-tf/Deblur_Net.py:79) = Add[T=DT_FLOAT, _device="/job:localhost/replica:0/task:0/device:GPU:0"](generator/Tanh-0-0-TransposeNCHWToNHWC-LayoutOptimizer, sub)]]

Does it only test the input size (1280, 720) ? Thanks!

dongheehand commented 5 years ago

I found an error if the test input size is not proper. So I will fix it soon If you want test on your own images, It seems better to crop the image properly.

SuamiL commented 5 years ago

have the same issue

SuamiL commented 5 years ago

oh I know why. when only test, don not add the "--chop_forward True", it will crop the image recursively and finally cause the size bug in some input size (for it does some computing work like "w/2 h/2 w-wc h-hc" and because of the python rounding rule, it may cause the difference). I haven't figure out how to fix the function but let --chop_forward be false can solve the issue