Minor changes to train_acd.py & processing.py

azure/train_acd.py

@@ -29,7 +29,7 @@
 
 LR = args.learning_rate
 WEIGHT = args.weight
-BANDS = ['B2', 'B3', 'B3', 'B8', 'B2_1', 'B3_1', 'B4_1', 'B8_1']
+BANDS = ['B2', 'B3', 'B4', 'B8', 'B2_1', 'B3_1', 'B4_1', 'B8_1']
 RESPONSE = args.response
 OPTIMIZER = tf.keras.optimizers.Adam(learning_rate=LR, beta_1=0.9, beta_2=0.999)
 LOSS = model.weighted_bce(WEIGHT)
@@ -66,8 +66,8 @@
         )
 
 # create special folders './outputs' and './logs' which automatically get saved
-os.makedirs('outputs', exists_OK = True)
-os.makedirs('logs', exists_OK = True)
+os.makedirs('outputs', exist_ok = True)
+os.makedirs('logs', exist_ok = True)
 out_dir = './outputs'
 log_dir = './logs'
 
@@ -86,12 +86,14 @@
     )
 
 # define a tensorboard callback to write training logs
-tensorboard = tf.keras.callbacks.Tensorboard(log_dir = log_dir)
+tensorboard = tf.keras.callbacks.TensorBoard(log_dir = log_dir)
 
 # train the model
-m.train(
+m.fit(
         x = training,
         epochs = args.epochs,
+        #TODO: make command line argument for size
+        steps_per_epoch = int(63*16),
         validation_data = evaluation,
         callbacks = [checkpoint, tensorboard]
         )

utils/processing.py

@@ -57,7 +57,7 @@ def augImg(img):
     outDims = tf.shape(img)[0:1]
     x = tf.image.random_flip_left_right(img)
     x = tf.image.random_flip_up_down(x)
-    x = tf.image.rot90(x, tf.random_uniform(shape=[], minval=0, maxval=4, dtype=tf.int32))
+    x = tf.image.rot90(x, tf.random.uniform(shape=[], minval=0, maxval=4, dtype=tf.int32))
       #x = zoom(x, outDims)
       #since were gonna map_fn this on a 4d image, output must be 3d, so squeeze the artificial 'sample' dimension
     return tf.squeeze(x)