ROS2 object detector using CenterNet
- nvidia-driver >= 396
- Ubuntu 16.04 or 18.04
- docker
- nvidia-docker2
- Ubuntu 18.04
- ROS2 (tested on dashing)
- See CenterNet for other requirement
- Published topic
- /detection_image (sensor_msgs/Image)
- image with bounding boxes and class name
- /bounding_boxes (std_msgs/String)
- json string including bounding_boxes information
- sample -> "{'list': [{'class': 'person', 'confidence': 0.898, 'xmin': 270.3, 'ymin': 12.6, 'xmax': 311.9, 'ymax': 39.0}, {'class': ...}, ...]}"
- /detection_image (sensor_msgs/Image)
- Subscribed topic
- /usb_cam/image_raw (sensor_msgs/Image)
cd your_workspace
git clone https://github.com/amslabtech/centernet_ros2.git
cd centernet_ros2
./build.sh
cd your_workspace/centernet_ros2
./run_docker.sh
(in docker container)
cd /root/ros2_ws/src/centernet_ros2/docker
./make_in_docker.sh
cd /root/ros2_ws
colcon build --symlink-install
source install/local_setup.bash
ros2 run centernet_ros2 object_detector