Update app.py

app.py

@@ -65,25 +65,16 @@ def inference_point(input_img, evt: gr.SelectData,):
     x, y = evt.index[0], evt.index[1]
     points = [[x, y]]
     print(f"Selected point: {points}")
-    import time
-    start_time = time.time()
     mp.set_start_method("spawn", force=True)
     config_file = './configs/ground-truth-warmup/mask-adapter/mask_adapter_convnext_large_cocopan_eval_ade20k.yaml'
     cfg = setup_cfg(config_file)
     
     demo = SAMPointVisualizationDemo(cfg, 0.8, sam2_model, clip_model,mask_adapter)
-    end_time = time.time()
-    print("init time",end_time - start_time)
     
-    start_time = time.time()
     img = read_image(input_img, format="BGR")
     
-    # Assume 'points' is a list of (x, y) coordinates to specify where the user clicks
-    # Process the image and points to create a segmentation map accordingly
     text_features = torch.from_numpy(np.load("./text_embedding/lvis_coco_text_embedding.npy"))
     _, visualized_output = demo.run_on_image_with_points(img, points,text_features)
-    end_time = time.time()
-    print("inf time",end_time - start_time)
     return visualized_output
 
 
@@ -171,8 +162,10 @@ with gr.Blocks() as demo:
         with gr.TabItem("Point Mode"):
             with gr.Row():  # 水平排列
                 with gr.Column(): 
+                    def init_state():
+                        return []
                     input_image = gr.Image(type='filepath', label="Upload Image", interactive=True)  # 上传图片并允许交互
-                    points_input = gr.State(value=[])  # 用于存储点击的点
+                    points_input = gr.State(value=init_state())  # 用于存储点击的点
 
                 with gr.Column():  # 第二列：分割图输出
                     output_image_point = gr.Image(type="pil", label='Segmentation Map')  # 输出分割图