Update app.py

app.py

@@ -11,7 +11,8 @@ model = AutoModelForCausalLM.from_pretrained("Qwen/Qwen-VL-Chat-Int4", device_ma
 
 def generate_predictions(image_input, text_input, with_grounding):
     user_image_path = "/tmp/user_input_test_image.jpg"
-    Image.fromarray((255 - (image_input * 255).astype('uint8'))).save(user_image_path)
+    original_image = Image.fromarray((255 - (image_input * 255).astype('uint8')))
+    original_image.save(user_image_path)
     
     if with_grounding == "Yes":
         text_input += " with grounding"
@@ -31,6 +32,10 @@ def generate_predictions(image_input, text_input, with_grounding):
     
     image_with_boxes = tokenizer.draw_bbox_on_latest_picture(full_response)
     
+    # Check if the response contains bounding box coordinates
+    if not re.search(r'\(\d+,\d+\),\(\d+,\d+\)', frontend_response):
+        image_with_boxes = original_image
+    
     if image_with_boxes:
         temp_path = "/tmp/image_with_boxes.jpg"
         image_with_boxes.save(temp_path)