Scenario runner

DatasetSave.py

@@ -27,6 +27,8 @@ def __init__(self, config, scenario_name):
         self.DEPTH_PATH = None 
         self.SEMANTIC_PATH = None
 
+        self.NEW_VIEW_PATH = None
+
         self.generate_path(self.config["SAVE_CONFIG"]["ROOT_PATH"], scenario_name)
         self.captured_frame_no = self.get_current_files_num()
 
@@ -44,7 +46,8 @@ def generate_path(self, root_path, scenario_name):
         self.OUTPUT_FOLDER = os.path.join(root_path, PHASE)
         folders = ['calib', 'image', 'image_label', 'bbox_img', 
               'velodyne', 'lidar_label', 'ego_state', 'extrinsic', 
-              'depth', 'semantic']
+              'depth', 'semantic', 'new_view', 'new_view/label',
+              'new_view/new_bbox']
 
         for folder in folders:
             directory = os.path.join(self.OUTPUT_FOLDER, folder)
@@ -65,7 +68,8 @@ def generate_path(self, root_path, scenario_name):
         self.EXTRINSIC_TXT_PATH = os.path.join(self.OUTPUT_FOLDER, 'extrinsic/{0:06}.txt')
         self.DEPTH_PATH = os.path.join(self.OUTPUT_FOLDER, 'depth/{0:06}_depth_{1}.png')
         self.SEMANTIC_PATH = os.path.join(self.OUTPUT_FOLDER, 'semantic/{0:06}_semantic_{1}.png')
-
+        self.NEW_VIEW_PATH = os.path.join(self.OUTPUT_FOLDER, 'new_view/{0:06}_new_view_{1}.png')
+        self.NEW_BBOX_PATH = os.path.join(self.OUTPUT_FOLDER, 'new_view/new_bbox/{0:06}_new_bbox_{1}.png')
 
 
     def get_current_files_num(self):
@@ -169,6 +173,23 @@ def save_datasets(self, data):
             for i in range(5)
         ]
 
+        # 新视角图片路径生成
+        new_view_files = [
+            self.NEW_VIEW_PATH.format(self.captured_frame_no, i)
+            for i in range(5)
+        ]
+
+         # 新视角标签路径生成（添加以下代码）
+        new_view_label_files = [
+            os.path.join(self.OUTPUT_FOLDER, f'new_view/label/{self.captured_frame_no:06}_new_view_{i}.txt')
+            for i in range(5)
+        ]
+
+        new_bbox_files = [
+            self.NEW_BBOX_PATH.format(self.captured_frame_no, i)
+            for i in range(5)
+        ]
+
         # 其他路径保持不变
         calib_filename = self.CALIBRATION_PATH.format(self.captured_frame_no)
         lidar_label_filename = self.LIDAR_LABEL_PATH.format(self.captured_frame_no)
@@ -202,13 +223,27 @@ def save_datasets(self, data):
                     dt["sensor_data"][[8, 9, 10, 15, 16][seg_cam]]
                 )
 
-            # 激光雷达数据保存（保持原始TODO注释）
-            save_lidar_data(lidar_files[0], dt["sensor_data"][2], extrinsic[2])
-            save_lidar_data(lidar_files[1], dt["sensor_data"][19], extrinsic[19])
-            save_lidar_data(lidar_files[2], dt["sensor_data"][20], extrinsic[20])
-            save_lidar_data(lidar_files[3], dt["sensor_data"][21], extrinsic[21])
-            save_lidar_data(lidar_files[4], dt["sensor_data"][22], extrinsic[22])
-            # save_lidar_data(lidar_files[1], dt["sensor_data"][6], extrinsic[6])  # 保持注释状态
+            # 新视角图像保存
+            for i, view_idx in enumerate([6, 23, 24, 25, 26]): 
+                save_image_data(new_view_files[i], dt["sensor_data"][view_idx])
+                # 保存对应标签
+                save_kitti_label_data(
+                    new_view_label_files[i],
+                    dt[f"image_labels_kitti_{5+i}"]  # 假设新增的5个标签存储在5-9索引
+                )
+                save_bbox_image_data(
+                    new_bbox_files[i],
+                    dt[f"bbox_img_{5+i}"]  # 假设bbox_img_5到9对应新视角
+                )
+            # save_bbox_image_data(new_bbox_files[1], dt["bbox_img_5"])
+            # print(new_bbox_files[0])
+
+            # 激光雷达数据保存
+            sensor_indices = [2, 19, 20, 21, 22]
+            for i, sensor_idx in enumerate(sensor_indices):
+                save_lidar_data(lidar_files[i], 
+                              dt["sensor_data"][sensor_idx], 
+                              extrinsic[sensor_idx])
             save_kitti_label_data(lidar_label_filename, dt["pc_labels_kitti"])
 
             # 保存标签和标注数据
@@ -221,6 +256,9 @@ def save_datasets(self, data):
                     bbox_img_filenames[cam_idx],
                     dt[f"bbox_img{'_'+str(cam_idx) if cam_idx>0 else ''}"]
                 )
+
+            # 将新视角的label保存到新视角的文件夹
+
             # EGO状态保存保持不变
             save_ego_data(
                 ego_state_filename,
@@ -234,7 +272,7 @@ def save_datasets(self, data):
             # # 保存深度和语义数据
             # for i, depth_idx in enumerate([1, 11, 12]):
             #     save_depth_image_data(depth_files[i], dt["sensor_data"][depth_idx])
-            # for i, seg_idx in enumerate([8, 9, 10]):
+            # for i, seg_idx in enumerate(32， 33， 34， 35， 36]):
             #     save_semantic_image_data(semantic_files[i], dt["sensor_data"][seg_idx])
 
         self.captured_frame_no += 1

SimulationScene.py

@@ -41,78 +41,23 @@ def __init__(self, config):
         self.random_seed = 17  # 固定随机种子
         random.seed(self.random_seed)
         np.random.seed(self.random_seed)
-        # 新增场景运行器实例
-        self.scenario_runner = None
-        self._init_scenario_runner()
-
-
-
-    def _init_scenario_runner(self):
-        """初始化场景运行器参数"""
-        self.args = argparse.Namespace(
-            host='127.0.0.1',
-            port=2000,
-            timeout=10.0,
-            trafficManagerPort=8001,
-            trafficManagerSeed=0,
-            sync=True,
-            scenario=self.config.get("SCENARIO", "FollowLeadingVehicleWithObstacle_2"),
-            openscenario=None,
-            openscenarioparams=None,
-            route=None,
-            agent='autopilot_agent.py',
-            agentConfig='',
-            output=True,
-            file=False,
-            junit=False,
-            json=False,
-            outputDir='',
-            configFile='',
-            additionalScenario=None,
-            debug=False,
-            reloadWorld=True,
-            record=None,
-            randomize=False,
-            repetitions=1,
-            waitForEgo=True,
-            # 以下参数在help中列出但未在代码中使用
-            version=False,
-            list=False,
-            osc=False
-        )
-        self.client = carla.Client(self.args.host, self.args.port)
-
-        # self.scenario_runner = ScenarioRunner(self.args)
-        print("Init")
-
-    def run_scenario(self):
-        """执行场景"""
-        try:
-            success = self.scenario_runner.run()
-            print(f"场景执行 {'成功' if success else '失败'}")
-
-            time.sleep(5)  # 延长等待时间至5秒
-            self.world = self.client.get_world()  # 刷新world对象
-
-            return success
-        except Exception as e:
-            logging.error(f"场景执行异常: {str(e)}")
-            return False
-        finally:
-            self.scenario_runner.destroy()
 
     def set_map(self):
         """
             设置场景地图
         """
         self.world = self.client.get_world()
 
-
     def set_weather(self):
         """
             设置场景天气
         """
-        pass
+        weather = carla.WeatherParameters(
+            cloudiness=10.0,
+            precipitation=0.0,
+            sun_altitude_angle=70.0
+        )
+        self.world.set_weather(weather)
 
     def set_synchrony(self):
         """

configs.yaml

@@ -57,15 +57,15 @@ SENSOR_CONFIG:
   #     location: [0, 0, 1.0]
   #     rotation: [0, 0, 0]
 
-  VIEW_RGB_1:
+  NEW_RGB_1:
     BLUEPRINT: sensor.camera.rgb
     ATTRIBUTE:
       image_size_x: 960
       image_size_y: 640
       fov: 90
     TRANSFORM:
-      location: [1.0, -3.2, 1.6]
-      rotation: [0, 30, 0]
+      location: [1.5, 0, 1.6]
+      rotation: [0, 0, 0]
 
   VIEW_RGB_BEV:
     BLUEPRINT: sensor.camera.rgb
@@ -162,6 +162,145 @@ SENSOR_CONFIG:
     BLUEPRINT: sensor.lidar.ray_cast
     ATTRIBUTE: {range: 70, rotation_frequency: 100, upper_fov: 20, lower_fov: -10, points_per_second: 20480000, channels: 128}
 
+  NEW_RGB_2:
+      BLUEPRINT: sensor.camera.rgb
+      ATTRIBUTE:
+        image_size_x: 960
+        image_size_y: 640
+        fov: 90
+      TRANSFORM:
+        location: [1.0, 3.2, 1.6]
+        rotation: [0, 0, 0]
+
+  NEW_RGB_3:
+      BLUEPRINT: sensor.camera.rgb
+      ATTRIBUTE:
+        image_size_x: 960
+        image_size_y: 640
+        fov: 90
+      TRANSFORM:
+        location: [1.0, 3.2, 1.6]
+        rotation: [0, 15, 0]
+
+  NEW_RGB_4:
+      BLUEPRINT: sensor.camera.rgb
+      ATTRIBUTE:
+        image_size_x: 960
+        image_size_y: 640
+        fov: 90
+      TRANSFORM:
+        location: [1.0, -3.2, 1.6]
+        rotation: [0, 0, 0]
+
+  NEW_RGB_5:
+    BLUEPRINT: sensor.camera.rgb
+    ATTRIBUTE:
+      image_size_x: 960
+      image_size_y: 640
+      fov: 90
+    TRANSFORM:
+      location: [1.0, 3.2, 1.6]
+      rotation: [0, -15, 0]
+
+  NEW_DEPTH_1:
+      BLUEPRINT: sensor.camera.depth
+      ATTRIBUTE:
+        image_size_x: 960
+        image_size_y: 640
+        fov: 90
+      TRANSFORM:
+        location: [1.5, 0, 1.6]
+        rotation: [0, 0, 0]
+
+  NEW_DEPTH_2:
+      BLUEPRINT: sensor.camera.depth
+      ATTRIBUTE:
+        image_size_x: 960
+        image_size_y: 640
+        fov: 90
+      TRANSFORM:
+        location: [1.0, 3.2, 1.6]
+        rotation: [0, 0, 0]
+
+  NEW_DEPTH_3:
+      BLUEPRINT: sensor.camera.depth
+      ATTRIBUTE:
+        image_size_x: 960
+        image_size_y: 640
+        fov: 90
+      TRANSFORM:
+        location: [1.0, 3.2, 1.6]
+        rotation: [0, 15, 0]
+
+  NEW_DEPTH_4:
+      BLUEPRINT: sensor.camera.depth
+      ATTRIBUTE:
+        image_size_x: 960
+        image_size_y: 640
+        fov: 90
+      TRANSFORM:
+        location: [1.0, -3.2, 1.6]
+        rotation: [0, 0, 0]
+
+  NEW_DEPTH_5:
+    BLUEPRINT: sensor.camera.depth
+    ATTRIBUTE:
+      image_size_x: 960
+      image_size_y: 640
+      fov: 90
+    TRANSFORM:
+      location: [1.0, 3.2, 1.6]
+      rotation: [0, -15, 0]
+
+  NEW_SEM_1:
+      BLUEPRINT: sensor.camera.semantic_segmentation
+      ATTRIBUTE:
+        image_size_x: 960
+        image_size_y: 640
+        fov: 90
+      TRANSFORM:
+        location: [1.5, 0, 1.6]
+        rotation: [0, 0, 0]
+
+  NEW_SEM_2:
+      BLUEPRINT: sensor.camera.semantic_segmentation
+      ATTRIBUTE:
+        image_size_x: 960
+        image_size_y: 640
+        fov: 90
+      TRANSFORM:
+        location: [1.0, 3.2, 1.6]
+        rotation: [0, 0, 0]
+
+  NEW_SEM_3:
+      BLUEPRINT: sensor.camera.semantic_segmentation
+      ATTRIBUTE:
+        image_size_x: 960
+        image_size_y: 640
+        fov: 90
+      TRANSFORM:
+        location: [1.0, 3.2, 1.6]
+        rotation: [0, 15, 0]
+
+  NEW_SEM_4:
+      BLUEPRINT: sensor.camera.semantic_segmentation
+      ATTRIBUTE:
+        image_size_x: 960
+        image_size_y: 640
+        fov: 90
+      TRANSFORM:
+        location: [1.0, -3.2, 1.6]
+        rotation: [0, 0, 0]
+
+  NEW_SEM_5:
+    BLUEPRINT: sensor.camera.semantic_segmentation
+    ATTRIBUTE:
+      image_size_x: 960
+      image_size_y: 640
+      fov: 90
+    TRANSFORM:
+      location: [1.0, 3.2, 1.6]
+      rotation: [0, -15, 0]
 
 FILTER_CONFIG:
   PRELIMINARY_FILTER_DISTANCE: 50
@@ -172,5 +311,5 @@ FILTER_CONFIG:
 
 SAVE_CONFIG:
   ROOT_PATH: data
-  STEP: 3
+  STEP: 10
   MAX_RECORD_COUNT: 200

main.py

@@ -56,7 +56,7 @@ def main():
 
         # 新增初始化等待阶段
         print("初始化完成，开始预运行...")
-        INIT_WAIT_FRAMES = 10  # 等待100帧（约5秒，假设20FPS）
+        INIT_WAIT_FRAMES = 30  # 等待100帧（约5秒，假设20FPS）
         for _ in range(INIT_WAIT_FRAMES):
             scene.update_spectator()  # 保持视角更新
             scene.world.tick()

run.sh

@@ -1,6 +1,6 @@
 #!/bin/bash
 
-SCENARIO_NAME="DynamicObjectCrossing_4"
+SCENARIO_NAME="FollowLeadingVehicleWithObstacle_1"
 # MAIN_PY_ARGS="--no-save" 
 MAIN_PY_ARGS=""