return inputs

app.py

@@ -284,12 +284,14 @@ def infer(
         )
     )
 
-    room_image_cropped = room_image.crop((
-        mask_bbox_x_min,
-        mask_bbox_y_min,
-        mask_bbox_x_max,
-        mask_bbox_y_max,
-    ))
+    room_image_cropped = room_image.crop(
+        (
+            mask_bbox_x_min,
+            mask_bbox_y_min,
+            mask_bbox_x_max,
+            mask_bbox_y_max,
+        )
+    )
     room_image_padded, room_image_padded_size = pad(
         room_image_cropped,
         (max_dimension, max_dimension),
@@ -311,12 +313,14 @@ def infer(
         ImageFilter.GaussianBlur(radius=kernel_size_blur)
     )
 
-    room_mask_cropped = room_mask_blurred.crop((
-        mask_bbox_x_min,
-        mask_bbox_y_min,
-        mask_bbox_x_max,
-        mask_bbox_y_max,
-    ))
+    room_mask_cropped = room_mask_blurred.crop(
+        (
+            mask_bbox_x_min,
+            mask_bbox_y_min,
+            mask_bbox_x_max,
+            mask_bbox_y_max,
+        )
+    )
     room_mask_padded, _ = pad(
         room_mask_cropped,
         (max_dimension, max_dimension),
@@ -372,6 +376,9 @@ def infer(
     )["images"]
 
     final_images = []
+    final_images.append(image)
+    final_images.append(mask)
+
     for image in results_images:
         final_image = room_image.copy()