Upload app.py with huggingface_hub

app.py

@@ -223,7 +223,11 @@ def run_spatial(image, task, question, options_text, quantity, unit, progress=gr
 
     prediction = result["prediction"]
     if task == "point":
-        point = prediction[0] if isinstance(prediction, list) else prediction
+        # decode_point returns [x, y] (a single point); visual-token variant returns a list of them
+        if isinstance(prediction[0], (list, tuple)):
+            point = prediction[0]
+        else:
+            point = prediction
         overlay = draw_point(img, point)
     else:
         overlay = img.copy()