snapshot...

modules/deeplearning/cloud_fraction_fcn_abi_hkm_refl.py

@@ -519,6 +519,7 @@ class SRCNN:
         momentum = 0.99
 
         num_filters = 64
+        num_filters_hkm = 8
 
         input_2d = self.inputs[0]
         input_hkm_2d = self.inputs[1]
@@ -530,12 +531,11 @@ class SRCNN:
         conv_hkm = conv_hkm_b = tf.keras.layers.Conv2D(num_filters, kernel_size=KERNEL_SIZE, kernel_initializer='he_uniform', activation=activation, padding='VALID')(input_hkm_2d)
         print(conv_hkm.shape)
 
-        num_filters_hkm = 8
         conv_hkm = build_conv2d_block(conv_hkm, num_filters_hkm, activation, 'Conv2D_1')
         num_filters_hkm *= 2
         conv_hkm = build_conv2d_block(conv_hkm, num_filters_hkm, activation, 'Conv2D_2')
 
-        conv_b = conv_b + conv_hkm
+        conv_b = tf.keras.layers.concatenate([conv_b, conv_hkm])
         print(conv_b.shape)
 
         # if NOISE_TRAINING:
@@ -549,7 +549,7 @@ class SRCNN:
 
         conv_b = build_residual_conv2d_block(conv_b, num_filters, 'Residual_Block_3', kernel_size=KERNEL_SIZE, scale=scale)
 
-        conv_b = build_residual_conv2d_block(conv_b, num_filters, 'Residual_Block_4', kernel_size=KERNEL_SIZE, scale=scale)
+        #conv_b = build_residual_conv2d_block(conv_b, num_filters, 'Residual_Block_4', kernel_size=KERNEL_SIZE, scale=scale)
 
         #conv_b = build_residual_conv2d_block(conv_b, num_filters, 'Residual_Block_5', kernel_size=KERNEL_SIZE, scale=scale)