SRCNN-Tensorflow
SRCNN-Tensorflow copied to clipboard
tegg89
Why is the size of "test_image" different from the original picture
The size of the "test_image" is 231231 while the size of the original "butterfly_GT" is 256256.
So when I compute the PSNR with matlab ,it's only about 11 dB, which is too small. I can't solve this problem, who can help me?
One more question ,How did you compute the PSNR? Did the author give the method of computing PSNR in this code? I can't find it.
First it's because the bottom row and further right set of patches is not computed. Use compare_psnr from skimage.measure using the original image cropped to the same size as test image. It should give u a better number.
On Sat, May 12, 2018, 3:00 AM MBCmbc [email protected] wrote:
One more question ,How did you compute the PSNR? Did the author give the method of coputing PSNR in this code? I can't find it.
— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/tegg89/SRCNN-Tensorflow/issues/29#issuecomment-388535415, or mute the thread https://github.com/notifications/unsubscribe-auth/ATogiPVWayXMcYZ9OhHafs8Ct4Y219laks5txohygaJpZM4T8Tjb .
I set the 'is_train' to False, but I find the size of 'test_image.png' is 336*336 and it seems the test image resolution is very low . Can anyone help me?
hi.. test image size is reduced as well as image is not same as the org butterfly image bcoz test image in sample is downscaled by a factor of 3 and then upscaled using the same factor. so that image is basically a low resolution image. but can u help me in knowing where the actual results are being stored after training and testing is completed
Only a single image is being tested in testing, refer to line 164 (preprocess() call in def input_setup). data[2] takes only the third image in "Set5" directory and results the output image as test_image.png in sample directory.
