junyanz / pytorch-CycleGAN-and-pix2pix

Image-to-Image Translation in PyTorch
Other
22.8k stars 6.29k forks source link

Testing Sat2map #1532

Closed reyntors closed 1 year ago

reyntors commented 1 year ago

I trained the model and I test my own aerial image

Why the generate result is bad, here is my output of my own aerial image test9_real_A test9_fake_B

Then i try to feed train images real_A in your datasets images:

test12_real_A test12_fake_B

What is the problem of this testing is there any problem in testing? Or please help me how to proper output. Thank you

Bea07 commented 1 year ago

check net.eval() function.. disable it and you get better results

junyanz commented 1 year ago

You may also need to train a new model on your own aerial images, as different aerial images have different data distributions.