update app.py

app.py

@@ -151,23 +151,25 @@ def generate(
         image_start_index_list=image_start_index_list,
         image_nums=image_nums,
     )
+
     boxes = outputs["boxes"]
     scores = outputs["scores"]
     if len(scores) > 0:
         box = boxes[scores.argmax()]
-    else:
-        iou = 0.0
-        # tqdm.write(f"output: {tokenizer.batch_decode(outputs)}")
-        tqdm.write(f"no output for: {uniq_id}, {image_id}, {text}")
     print(f"{box}")
 
-
-    gen_text = tokenizer.batch_decode(outputs)
     if idx == 1:
-        return f"Output:{gen_text}", out_image
+        open_cv_image = np.array(image)
+        # Convert RGB to BGR
+        open_cv_image = open_cv_image[:, :, ::-1].copy()
+        open_cv_image = cv2.rectangle(open_cv_image, box[:2].astype(int), box[2:].astype(int), (255, 0, 0), 2)
+        out_image = Image.fromarray(cv2.cvtColor(open_cv_image, cv2.COLOR_BGR2RGB))
+        return f"Output:{box}", out_image
     elif idx == 2:
+        gen_text = tokenizer.batch_decode(outputs)
         return (f"Question: {text.strip()} Answer: {gen_text}")
     else:
+        gen_text = tokenizer.batch_decode(outputs)
         return (f"Output:{gen_text}")