Update app.py

app.py

@@ -19,6 +19,8 @@ TEST_FOLDER = 'example_images'
 NUM_CLASSES = 7
 
 def pil_image_as_numpy_array(pilimg):
+
+    # Convert PIL image to numpy array with Tensorflow utils function
     img_array = tf.keras.utils.img_to_array(pilimg)
     return img_array
     
@@ -77,17 +79,6 @@ def get_predictions(y_prediction_encoded):
   return predicted_label_indices
 
 def predict_on_train(image):
-
-    # Extract filename from Gradio input
-    image_filename = gr.get_data()[0].name
-    # Construct the filename for the ground truth mask
-    mask_filename = image_filename.replace('_sat.jpg', '_mask.png')
-
-    # Load the ground truth mask
-    mask_path = os.path.join(TRAIN_FOLDER, mask_filename)
-    ground_truth_mask = Image.open(mask_path)
-    # Resize the mask image
-    ground_truth_mask_pil = resize_image(ground_truth_mask)
     
     # Steps to get prediction of the satellite image
     sample_image_resized = resize_image(image)
@@ -121,7 +112,7 @@ def predict_on_train(image):
     # Close the figure to release resources
     plt.close(fig)
 
-    return ground_truth_mask_pil, image_pil
+    return image_pil, image_pil
 
 def predict_on_test(image):
     
@@ -174,15 +165,6 @@ description= '''
     validation accuracy of about 75% and dice score of about 0.6.
     '''
 
-# Launch Gradio Interface (Single Tab interface)
-# gr.Interface(
-#     predict,
-#     title='Land Cover Segmentation',
-#     inputs=[gr.Image()],
-#     outputs=[gr.Image()],
-#     examples=sample_images
-# ).launch(debug=True, share=True)
-
 # Create the train dataset interface
 tab1 = gr.Interface(
     fn=predict_on_train,