Updated app.py

app.py

@@ -58,6 +58,7 @@ def preprocess_image(image):
     - Normalize pixel values
     - Convert RGB to BGR if required by the model
     """
+    image = image.convert('RGB')  # Ensure image is in RGB
     image = image.resize((TARGET_WIDTH, TARGET_HEIGHT))
     image_np = np.array(image).astype(np.float32)
     image_np = image_np / 255.0  # Normalize to [0,1]
@@ -84,9 +85,12 @@ def draw_boxes(image, boxes, classes, scores, threshold=0.5):
     except IOError:
         font = ImageFont.load_default()
     
+    detections = False  # Flag to check if any detections are above threshold
+
     for box, cls, score in zip(boxes[0], classes[0], scores[0]):
         if score < threshold:
             continue
+        detections = True
         # Convert box coordinates from normalized to absolute
         ymin, xmin, ymax, xmax = box
         left = xmin * image.width
@@ -111,6 +115,30 @@ def draw_boxes(image, boxes, classes, scores, threshold=0.5):
         # Draw text
         draw.text((left + 2, top - text_height - 2), label, fill="white", font=font)
     
+    if not detections:
+        # Optionally, you can add text indicating no detections were found
+        try:
+            font_large = ImageFont.truetype("arial.ttf", 20)
+        except IOError:
+            font_large = ImageFont.load_default()
+        no_detect_text = "No detections found."
+        text_bbox = draw.textbbox((0, 0), no_detect_text, font=font_large)
+        text_width = text_bbox[2] - text_bbox[0]
+        text_height = text_bbox[3] - text_bbox[1]
+        draw.rectangle(
+            [
+                ((image.width - text_width) / 2 - 10, (image.height - text_height) / 2 - 10),
+                ((image.width + text_width) / 2 + 10, (image.height + text_height) / 2 + 10)
+            ],
+            fill="black"
+        )
+        draw.text(
+            ((image.width - text_width) / 2, (image.height - text_height) / 2),
+            no_detect_text,
+            fill="white",
+            font=font_large
+        )
+    
     return image
 
 def predict(image):
@@ -122,17 +150,25 @@ def predict(image):
         PIL.Image: Annotated image with bounding boxes and labels.
     """
     try:
+        print("Starting prediction...")
         # Preprocess the image
         input_array = preprocess_image(image)
+        print(f"Preprocessed image shape: {input_array.shape}")
         
         # Run inference
         boxes, classes, scores = sess.run(
             [detected_boxes, detected_classes, detected_scores],
             feed_dict={input_tensor: input_array}
         )
+        print(f"Inference completed. Number of detections: {len(boxes[0])}")
+        
+        # Check if detections are present
+        if boxes.shape[1] == 0:
+            print("No detections returned by the model.")
         
         # Annotate the image with bounding boxes and labels
         annotated_image = draw_boxes(image.copy(), boxes, classes, scores, threshold=0.5)
+        print("Annotated image created successfully.")
         
         return annotated_image