fixed train.py

train.py

@@ -1,23 +1,20 @@
 #!/usr/bin/env python
 from patched_cnn import *
 
-shadows = open_images("./segments/shadows")
-non_shadows = open_images("./segments/non_shadows", len(shadows))
+images = open_images("./data/SBUTrain4KRecoveredSmall/ShadowImages", 4070)
+shadow_masks = open_images("./data/SBUTrain4KRecoveredSmall/ShadowMasks", 4070, True)
 
 x = [] # input features.
 y = [] # labels
 
-x.extend(shadows)
-x.extend(non_shadows)
+x.extend(images)
+y.extend(shadow_masks)
 
-y.extend([ 1 for i in range(len(shadows)) ])
-y.extend([ 0 for i in range(len(non_shadows)) ])
-
-cnn = Patched_CNN()
-cnn.build_model()
-
-batch_size = 50
-epochs = 50
-patience = 5
-cnn.train(x, y, batch_size, epochs, patience)
-cnn.save_model()
+prior_cnn = Patched_CNN()
+prior_cnn.build_model(channels=3)
+prior_cnn.train(
+        x, y,
+        batch_size=20,
+        epochs=100,
+        patience=5,
+        prefix="complete_images_")