Update app.py

app.py

@@ -37,23 +37,24 @@ def recognize_handwritten_text(image):
             return Image.fromarray(processed_image), "No text detected"
         
         boxes = outputs[0]["boxes"]
-        print(f"Debug: Boxes structure = {boxes}")  # Debug output to logs
+        print(f"Debug: Boxes structure = {boxes}")  # Log the exact structure
         pil_image = Image.fromarray(processed_image)
         texts = []
         
-        # Adjust box unpacking based on actual structure
+        # Handle box format (assuming [x, y, width, height] or [[x1, y1], [x2, y2]])
         for box in boxes:
-            if len(box) == 4:  # [x, y, width, height] format
-                x_min, y_min, width, height = box
-                x_max = x_min + width
-                y_max = y_min + height
-            elif len(box) == 2 and len(box[0]) == 2:  # [[x, y], [width, height]] format
-                x_min, y_min = box[0]
-                width, height = box[1]
-                x_max = x_min + width
-                y_max = y_min + height
+            if len(box) == 4:  # [x, y, width, height]
+                x, y, width, height = box
+                x_min, y_min = x, y
+                x_max, y_max = x + width, y + height
+            elif len(box) == 2 and all(len(p) == 2 for p in box):  # [[x1, y1], [x2, y2]]
+                x1, y1 = box[0]
+                x2, y2 = box[1]
+                x_min, y_min = min(x1, x2), min(y1, y2)
+                x_max, y_max = max(x1, x2), max(y1, y2)
             else:
-                continue  # Skip invalid boxes
+                print(f"Debug: Skipping invalid box {box}")  # Log invalid boxes
+                continue
             
             crop = pil_image.crop((x_min, y_min, x_max, y_max))
             pixel_values = processor(images=crop, return_tensors="pt").pixel_values