Update app.py

app.py

@@ -66,12 +66,11 @@ def process_video(video_path, process_seconds=20, conf_threshold=0.2, max_det=8)
         (width, height)
     )
     
-    # 计算点的大小
-    # 基于视频分辨率计算合适的点大小
+    # 计算基于分辨率的点大小和线宽
     base_size = min(width, height)
-    point_size = max(1, int(base_size * 0.005))  # 0.5% 的最小边长作为点大小
-    line_thickness = max(1, int(base_size * 0.002))  # 0.2% 的最小边长作为线宽
-    
+    point_size = max(1, int(base_size * 0.005))  # 0.5% 的最小边长
+    line_thickness = max(1, int(base_size * 0.002))  # 0.2% 的最小边长
+
     # 设置推理参数并处理视频
     results = model.predict(
         source=video_path,
@@ -80,12 +79,14 @@ def process_video(video_path, process_seconds=20, conf_threshold=0.2, max_det=8)
         save=False,
         show=False,
         stream=True,
-        line_width=line_thickness,  # 使用计算出的线宽
+        line_width=line_thickness,  # 线宽
         show_boxes=True,
         show_labels=True,
         show_conf=True,
         vid_stride=1,
         max_det=max_det,
+        kpt_radius=point_size,  # 关键点半径
+        kpt_line_thickness=line_thickness,  # 关键点连接线宽度
     )
     
     # 处理结果
@@ -96,19 +97,6 @@ def process_video(video_path, process_seconds=20, conf_threshold=0.2, max_det=8)
         # 获取绘制了预测结果的帧
         frame = r.plot()
         
-        # 如果需要自定义绘制姿态点
-        if hasattr(r, 'keypoints') and r.keypoints is not None:
-            for kpts in r.keypoints:
-                for x, y, conf in kpts:
-                    if conf > conf_threshold:
-                        cv2.circle(
-                            frame, 
-                            (int(x), int(y)), 
-                            point_size,  # 使用计算出的点大小
-                            (0, 255, 0), 
-                            -1  # -1 表示填充圆
-                        )
-        
         # 收集检测信息
         frame_info = {
             "frame": frame_count + 1,
@@ -143,13 +131,11 @@ def process_video(video_path, process_seconds=20, conf_threshold=0.2, max_det=8)
 - 置信度阈值: {conf_threshold:.2f}
 - 最大检测数量: {max_det}
 - 处理时长: {process_seconds}秒
-
 分析结果:
 - 处理帧数: {frame_count}
 - 平均每帧检测到的老鼠数: {np.mean([info['count'] for info in detection_info]):.1f}
 - 最大检测数: {max([info['count'] for info in detection_info])}
 - 最小检测数: {min([info['count'] for info in detection_info])}
-
 置信度分布:
 {np.histogram([float(det['confidence'].strip('%'))/100 for info in detection_info for det in info['detections']], bins=5)[0].tolist()}
 """
@@ -189,7 +175,7 @@ with gr.Blocks() as demo:
                 max_det = gr.Slider(
                     minimum=1,
                     maximum=10,
-                    value=1,
+                    value=8,
                     step=1,
                     label="最大检测数量",
                     info="每帧最多检测的目标数量"