I have taken a look at your work. Overall, it is nice. However, I have some comments below:
Why don't you normalize the input image in your custom data set?
Is converting to grayscale needed for the mask in this case?
Otherwise, you had better remove the warning message in your code by removing the interpolation method of image resizing. It would be very cruel if there were messages pumped out each time you sample a pair of (x, y) during the training phase.
I have taken a look at your work. Overall, it is nice. However, I have some comments below:
Otherwise, you had better remove the warning message in your code by removing the interpolation method of image resizing. It would be very cruel if there were messages pumped out each time you sample a pair of (x, y) during the training phase.