boxmot_ros

Docker Usage by adding to compose.yml file

To use GPU with docker while on AMD64 systems, install nvidia-container-toolkit with given instructions.

Supported platforms

Replace image and 'device' parameter in the compose.yml with following values for respective systems.

System	ROS Version	Value for image	Value for device	Size	file
AMD64	Humble	ghcr.io/kalanaratnayake/boxmot-ros:humble	cpu, 0, 0,1,2	5.7 GB	docker/compose.amd64.yaml

Docker Usage with this repository

Clone this reposiotory

mkdir -p boxmot_ws/src && cd boxmot_ws/src
git clone https://github.com/KalanaRatnayake/boxmot_ros.git && cd ..

on AMD64

Pull the Docker image and start compose (No need to run docker compose build)

cd src/boxmot_ros/docker
docker compose -f compose.amd64.yaml pull
docker compose -f compose.amd64.yaml up

Reset the system and remove volume

docker compose -f compose.amd64.yaml down
docker volume rm docker_boxmot

Native Usage

Clone this repository with and install dependencies.

git clone https://github.com/KalanaRatnayake/boxmot_ros.git
git clone https://github.com/KalanaRatnayake/detection_msgs.git
pip3 install -r boxmot_ros/requirements.txt

cd ..
rosdep install --from-paths src -y --ignore-src

Build the package

If required, edit the parameters at `config/boxmot_ros_params.yaml' and then at the workspace root run,

colcon build

Start the system

To use the boxmot tracking, run the launch file in terminal, run,

source ./install/setup.bash
ros2 launch boxmot_ros boxmot.launch.py

Parameter decription

ROS Parameter	Docker ENV parameter	Default Value	Description
yolo_model	YOLO_MODEL	yolov9t.pt	Model to be used. see [1] for default models and [2] for custom models
tracking_model	TRACKING_MODEL	deepocsort	Model to be used for tracking. see [3] for default models
reid_model	REID_MODEL	osnet_x0_25_msmt17.pt	Model to be used for reidentification. see [3] for default models
subscribe_depth	SUBSCRIBE_DEPTH	False	Whether to subscribe to depth image or not. This will also enable the depth_topic variable which publishes synchronized depth image. Use if having a depth camera. A ApproximateTimeSynchronizer is used to sync RGB and Depth images
input_rgb_topic	INPUT_RGB_TOPIC	/camera/color/image_raw	Topic to subscribe for RGB image. Accepts sensor_msgs/Image
input_depth_topic	INPUT_DEPTH_TOPIC	/camera/depth/points	Topic to subscribe for Depth image. Accepts sensor_msgs/PointCloud2
publish_annotated_image	PUBLISH_ANNOTATED_IMAGE	False	Whether to publish tracking annotated image, increases callback execution time when set to True
rgb_topic	RGB_TOPIC	/boxmot_ros/rgb_image	Topic for publishing synchronized rgb images. uses sensor_msgs/Image
depth_topic	DEPTH_TOPIC	/boxmot_ros/depth_image	Topic for publishing synchronized depth images. uses detection_msgs/PointCloud2
annotated_topic	ANNOTATED_TOPIC	/boxmot_ros/annotated_image	Topic for publishing annotated images uses sensor_msgs/Image
detailed_topic	DETAILED_TOPIC	/boxmot_ros/tracking_result	Topic for publishing detailed results uses detection_msgs/Detections
threshold	THRESHOLD	0.25	Confidence threshold for predictions
device	DEVICE	'0'	cpu for CPU, 0 for gpu, 0,1,2,3 if there are multiple GPUs

[1] If the model is available at ultralytics models, It will be downloaded from the cloud at the startup. We are using docker volumes to maintain downloaded weights so that weights are not downloaded at each startup.

[2] Uncomment the commented out YOLO_MODEL parameter line and give the custom model weight file's name as YOLO_MODEL parameter. Uncomment the docker bind entry that to direct to the weights folder and comment the docker volume entry for yolo. Copy the custom weights to the weights folder.

[3] If the reid model is available at MODEL_ZOO, and tracking_model is supported [deepocsort, strongsort, ocsort, bytetrack, botsort, hybridsort, imprassoc]. They will be downloaded from the cloud at the startup. We are using docker volumes to maintain downloaded weights so that weights are not downloaded at each startup. Use the snipped in Default models with Docker Compose

Latency description

Here is a summary of whether supported models work with boxmot_ros node (in docker) on various platforms and the time it takes to execute a single interation of BoxmotROS.image_callback function. Values are measured as an average of 100 executions of the function and Input is a 640x480 RGB image at 30 fps.

Laptop -> Ryzen 9 16 core with RTX3070 mobile GPU with Ubuntu 22.04

Model	Laptop (amd64)
deepocsort	27 ms
strongsort	20 ms
ocsort	17 ms
bytetrack	17 ms
botsort	14 ms