pointpillar调用ROS实时检测KITTI.bag包

硬性要求：8—10G的显存（pointpillar运行刚需），可以用nvidia-smi查询
测试环境：ubuntu 18.04 CUDA 11.1 pytorch1.8.1
1、安装openpcdet运行环境
githut链接
2、安装ROS
我用的鱼香大佬的一键安装链接，第一次安装失败，不信邪第二次就直接成功了。
ROS安装链接
3、然后是编译环境，下载代码

mkdir -p ~/pointpillars_ros/src
cd pointpillars_ros/src
git clone https://github.com/BIT-DYN/pointpillars_ros
cd ..

4、下载相关库（其实好多库在鱼香大佬的一键安装ROS里面已经下载好了）

conda activate openpcdet # 激活openpcdet的环境
pip install --user rospkg catkin_pkg
pip install pyquaternion
sudo apt-get install ros-kinetic-pcl-ros
sudo apt-get install ros-kinetic-jsk-recognition-msgs # 之前找的一些相关帖子都是装的ros-kinetic-jsk-recognition-msg，然后就一直无法定位，等我去搜索这个库的时候好像是最后少了一个s，最后搞鼓半天发现已经安装好了
sudo apt-get install ros-kinetic-jsk-rviz-plugins
catkin_make # 注意编译的位置 在src同级目录

5、复制openpcdet的文件

我是把所有文件都复制下来了

6、修改其中的ros.py文件

sys.path.append("/home/optobot-03/pointpillars_ros-main/src/pointpillars_ros") # 第8行
config_path = rospy.get_param("/config_path", "/home/optobot-03/pointpillars_ros-main/src/pointpillars_ros/tools/cfgs/kitti_models/pointpillar.yaml")
ckpt_path = rospy.get_param("/ckpt_path", "/home/optobot-03/pointpillars_ros-main/src/pointpillars_ros/tools/models/pointpillar_7728.pth") # 第54、55行

7、修改.rviz文件
修改话题topic，如果是kitti可以直接抄我的：

Panels:
  - Class: rviz/Displays
    Help Height: 78
    Name: Displays
    Property Tree Widget:
      Expanded:
        - /Global Options1
        - /Status1
        - /PointCloud21
        - /BoundingBoxArray1
      Splitter Ratio: 0.5
    Tree Height: 235
  - Class: rviz/Selection
    Name: Selection
  - Class: rviz/Tool Properties
    Expanded:
      - /2D Pose Estimate1
      - /2D Nav Goal1
      - /Publish Point1
    Name: Tool Properties
    Splitter Ratio: 0.5886790156364441
  - Class: rviz/Views
    Expanded:
      - /Current View1
    Name: Views
    Splitter Ratio: 0.5
  - Class: rviz/Time
    Experimental: false
    Name: Time
    SyncMode: 0
    SyncSource: PointCloud2
Preferences:
  PromptSaveOnExit: true
Toolbars:
  toolButtonStyle: 2
Visualization Manager:
  Class: ""
  Displays:
    - Alpha: 0.5
      Cell Size: 1
      Class: rviz/Grid
      Color: 160; 160; 164
      Enabled: true
      Line Style:
        Line Width: 0.029999999329447746
        Value: Lines
      Name: Grid
      Normal Cell Count: 0
      Offset:
        X: 0
        Y: 0
        Z: 0
      Plane: XY
      Plane Cell Count: 10
      Reference Frame: <Fixed Frame>
      Value: true
    - Alpha: 1
      Autocompute Intensity Bounds: true
      Autocompute Value Bounds:
        Max Value: 26.235191345214844
        Min Value: -12.199440002441406
        Value: true
      Axis: Z
      Channel Name: intensity
      Class: rviz/PointCloud2
      Color: 255; 255; 255
      Color Transformer: FlatColor
      Decay Time: 0
      Enabled: true
      Invert Rainbow: false
      Max Color: 255; 255; 255
      Min Color: 0; 0; 0
      Name: PointCloud2
      Position Transformer: XYZ
      Queue Size: 10
      Selectable: true
      Size (Pixels): 3
      Size (m): 0.05000000074505806
      Style: Squares
      Topic: /kitti/velo/pointcloud
      Unreliable: false
      Use Fixed Frame: true
      Use rainbow: true
      Value: false
    - Class: rviz/Axes
      Enabled: true
      Length: 1
      Name: Axes
      Radius: 0.10000000149011612
      Reference Frame: <Fixed Frame>
      Value: true
      Topic: /modified
      Unreliable: false
      Use Fixed Frame: true
      Use rainbow: true
      Value: true
    - Class: jsk_rviz_plugin/BoundingBoxArray
      Enabled: true
      Name: BoundingBoxArray
      Topic: /detections
      Unreliable: false
      Value: true
      Queue Size: 10
      alpha: 0.800000011920929
      color: 25; 255; 0
      coloring: Flat color
      line width: 0.20000000298023224
      only edge: true
      show coords: false
    - Class: rviz/Image
      Enabled: true
      Image Topic: /kitti/camera_color_left/image_raw
      Max Value: 1
      Median window: 5
      Min Value: 0
      Name: Image
      Normalize Range: true
      Queue Size: 2
      Transport Hint: raw
      Unreliable: false
      Value: true
  Enabled: true
  Global Options:
    Background Color: 48; 48; 48
    Default Light: true
    Fixed Frame: velo_link
    Frame Rate: 30
  Name: root
  Tools:
    - Class: rviz/Interact
      Hide Inactive Objects: true
    - Class: rviz/MoveCamera
    - Class: rviz/Select
    - Class: rviz/FocusCamera
    - Class: rviz/Measure
    - Class: rviz/SetInitialPose
      Theta std deviation: 0.2617993950843811
      Topic: /initialpose
      X std deviation: 0.5
      Y std deviation: 0.5
    - Class: rviz/SetGoal
      Topic: /move_base_simple/goal
    - Class: rviz/PublishPoint
      Single click: true
      Topic: /clicked_point
  Value: true
  Views:
    Current:
      Class: rviz/Orbit
      Distance: 61.30393600463867
      Enable Stereo Rendering:
        Stereo Eye Separation: 0.05999999865889549
        Stereo Focal Distance: 1
        Swap Stereo Eyes: false
        Value: false
      Focal Point:
        X: 11.016645431518555
        Y: -6.342355728149414
        Z: 3.211975574493408
      Focal Shape Fixed Size: true
      Focal Shape Size: 0.05000000074505806
      Invert Z Axis: false
      Name: Current View
      Near Clip Distance: 0.009999999776482582
      Pitch: 1.124796748161316
      Target Frame: <Fixed Frame>
      Value: Orbit (rviz)
      Yaw: 3.210404872894287
    Saved: ~
Window Geometry:
  Displays:
    collapsed: false
  Height: 1011
  Hide Left Dock: false
  Hide Right Dock: false
  Image:
    collapsed: false
  QMainWindow State: 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
  Selection:
    collapsed: false
  Time:
    collapsed: false
  Tool Properties:
    collapsed: false
  Views:
    collapsed: false
  Width: 1920
  X: 0
  Y: 38

参考链接
8、修改.launch文件
也是参考上方这个链接

<launch>
  <node pkg="rosbag" type="play" name="player" output="log" args="-l /media/optobot-03/document/1/2022-10-15-15-01-20.bag" /> # 只需要改这一行就行
  <node name="pointpillars_ros" type="ros.py" pkg="pointpillars_ros" output="screen"/>
  <node type="rviz" name="rviz" pkg="rviz" args="-d $(find pointpillars_ros)/launch/pointpillars.rviz" />
</launch>

9、运行
在src目录的上一级运行

conda activate openpcdet
source ~/pointpillars_ros/devel/setup.bash
roslaunch pointpillars_ros pointpillars.launch

注意把这里改成/modified

10、结果
个人感觉这个包跑的有点快，我又根据这个bag包自己录了个0.5倍速的包，用的指令 record my desktop.