Update app.py

app.py

@@ -23,6 +23,7 @@ processor = Owlv2Processor.from_pretrained(
 @spaces.GPU
 def query_image(img, text_queries, score_threshold):
 
+    # Split queries (still required internally)
     text_queries = text_queries.split(",")
 
     # Prepare inputs
@@ -35,14 +36,15 @@ def query_image(img, text_queries, score_threshold):
     with torch.no_grad():
         outputs = model(**inputs)
 
-    # Move to CPU for processing
+    # Move outputs to CPU
     outputs.logits = outputs.logits.cpu()
     outputs.pred_boxes = outputs.pred_boxes.cpu()
 
-    # Correct target size
+    # Correct target size (IMPORTANT)
     target_sizes = torch.tensor([img.shape[:2]])
 
-    results = processor.post_process_object_detection(
+    # ✅ FIXED METHOD (important!)
+    results = processor.post_process_grounded_object_detection(
         outputs=outputs,
         target_sizes=target_sizes
     )[0]
@@ -52,6 +54,7 @@ def query_image(img, text_queries, score_threshold):
 
     output_boxes = []
 
+    # Process detections
     for box, score in zip(boxes, scores):
 
         if score < score_threshold:
@@ -59,10 +62,10 @@ def query_image(img, text_queries, score_threshold):
 
         x1, y1, x2, y2 = map(int, box.tolist())
 
-        # store ONLY coordinates
+        # Save ONLY coordinates
         output_boxes.append([x1, y1, x2, y2])
 
-        # draw rectangle (optional)
+        # Draw rectangle ONLY (no labels)
         cv2.rectangle(img, (x1, y1), (x2, y2), (0, 255, 0), 2)
 
     return img, output_boxes
@@ -80,9 +83,10 @@ demo = gr.Interface(
     ],
     outputs=[
         gr.Image(label="Bounding Boxes"),
-        gr.JSON(label="Coordinates")
+        gr.JSON(label="Coordinates Only")
     ],
     title="OWLv2 Bounding Box Coordinates Only"
 )
 
+# Launch app
 demo.launch()