修改摄像头排布,新增显示窗口合并为一个的函数进行窗口排布。

lerobot/common/robot_devices/control_utils.py

@@ -260,10 +260,11 @@ def control_loop(
             dataset.add_frame(frame)
 
         if display_cameras and not is_headless():
-            image_keys = [key for key in observation if "image" in key]
-            for key in image_keys:
-                cv2.imshow(key, cv2.cvtColor(observation[key].numpy(), cv2.COLOR_RGB2BGR))
-            cv2.waitKey(1)
+            # image_keys = [key for key in observation if "image" in key]
+            # for key in image_keys:
+            #     cv2.imshow(key, cv2.cvtColor(observation[key].numpy(), cv2.COLOR_RGB2BGR))
+            # cv2.waitKey(1)
+            display_observations_combined(observation)
 
         if fps is not None:
             dt_s = time.perf_counter() - start_loop_t
@@ -277,6 +278,107 @@ def control_loop(
             events["exit_early"] = False
             break
 
+import numpy as np
+def display_observations_combined(observation, display_cameras=True):
+    """将摄像头画面组合在一个窗口中显示，优化以防止抖动"""
+    if display_cameras and not is_headless():
+        image_keys = [key for key in observation if "image" in key]
+        if not image_keys:
+            return
+            
+        # 获取所有图像并转换为BGR格式
+        images = []
+        for key in image_keys:
+            img = observation[key].numpy()
+            # 确保每个图像尺寸为640×480
+            if img.shape[0] != 480 or img.shape[1] != 640:
+                img = cv2.resize(img, (640, 480))
+            img = cv2.cvtColor(img, cv2.COLOR_RGB2BGR)
+            # 添加图像标题
+            cv2.putText(img, key, (10, 30), 
+                       cv2.FONT_HERSHEY_SIMPLEX, 0.7, (0, 255, 0), 2)
+            images.append(img)
+        
+        # 根据实际摄像头数量确定布局
+        num_cameras = len(images)
+        
+        # 为固定数量的摄像头定义固定布局
+        margin = 10  # 图像之间的间距
+        
+        if num_cameras == 1:
+            # 单个摄像头 - 直接显示
+            grid_image = images[0]
+        elif num_cameras == 2:
+            # 两个摄像头 - 水平排列
+            grid_width = 2 * 640 + margin
+            grid_height = 480
+            grid_image = np.zeros((grid_height, grid_width, 3), dtype=np.uint8)
+            
+            # 放置两个图像
+            grid_image[0:480, 0:640] = images[0]
+            grid_image[0:480, 640+margin:] = images[1]
+        elif num_cameras == 3:
+            # 三个摄像头 - 上面一行2个，下面一个
+            grid_width = 2 * 640 + margin
+            grid_height = 2 * 480 + margin
+            grid_image = np.zeros((grid_height, grid_width, 3), dtype=np.uint8)
+            
+            # 放置三个图像
+            grid_image[0:480, 0:640] = images[0]
+            grid_image[0:480, 640+margin:] = images[1]
+            grid_image[480+margin:, (grid_width-640)//2:(grid_width+640)//2] = images[2]  # 居中放置
+        else:
+            # 四个或更多摄像头 - 2×2网格
+            grid_cols = 2
+            grid_rows = 2
+            grid_width = grid_cols * 640 + (grid_cols - 1) * margin
+            grid_height = grid_rows * 480 + (grid_rows - 1) * margin
+            grid_image = np.zeros((grid_height, grid_width, 3), dtype=np.uint8)
+            
+            # 最多显示4个
+            for i, img in enumerate(images[:4]):
+                row = i // grid_cols
+                col = i % grid_cols
+                
+                y_start = row * (480 + margin)
+                y_end = y_start + 480
+                x_start = col * (640 + margin)
+                x_end = x_start + 640
+                
+                grid_image[y_start:y_end, x_start:x_end] = img
+        
+        # 创建一个固定名称的窗口
+        window_name = "Camera Views"
+        
+        # 使用 getWindowProperty 检查窗口是否已经存在
+        try:
+            # 尝试获取窗口属性 - 如果窗口不存在会抛出异常
+            window_exists = cv2.getWindowProperty(window_name, cv2.WND_PROP_VISIBLE) >= 0
+        except:
+            window_exists = False
+        
+        # 如果窗口不存在，创建并定位它
+        if not window_exists:
+            cv2.namedWindow(window_name, cv2.WINDOW_NORMAL)
+            
+            # 设置窗口位置和大小（只需要在创建时设置一次）
+            screen_width = 1920  # 您的屏幕宽度
+            screen_height = 1080  # 您的屏幕高度
+            
+            # 根据实际图像计算合适的显示尺寸
+            scale_factor = min(screen_width / grid_width, screen_height / grid_height) * 0.95
+            display_width = int(grid_width * scale_factor)
+            display_height = int(grid_height * scale_factor)
+            
+            cv2.resizeWindow(window_name, display_width, display_height)
+            cv2.moveWindow(window_name, (screen_width - display_width) // 2, 
+                          (screen_height - display_height) // 2)  # 居中显示
+        
+        # 显示图像
+        cv2.imshow(window_name, grid_image)
+        cv2.waitKey(1)
+
+
 
 def reset_environment(robot, events, reset_time_s, fps):
     # TODO(rcadene): refactor warmup_record and reset_environment