0

I'm feeding keras model with generator.flow_from_directory yield masked, [orig, masks]. So passing a list into y input. But I don't understand what shape y_true will take in this case. orig and masks are 4d arrays of shape [32, 256, 256, 3] so I thought it should be 5d tensor but it's not the case. I looked here but unfortunately solutions from there don't work. print(Y_train.shape) gives an error and model.summary() doesn't print anything. print(y_true.shape) gives (None, None, None, None). How do I understand the shape and how do I take original list elements from y_true separately?

Ivan Smirnov
  • 53
  • 8
  • Can you be more specific or what you're trying to solve ? – thushv89 Nov 16 '19 at 23:14
  • @thushv89 I have data (image masks) used in Input calculation (masked images). Maks are generated during training. I need to pass these masks to my custom loss function. So I thought I could do it with passing y to model.fit as a list of original images and masks. But I don't understand how to split y_true and to use separately images and masks in loss calculation. – Ivan Smirnov Nov 16 '19 at 23:23

0 Answers0