TRT YOLO

Support List:

Model Name	size	task	Real
Yolov8	n	detection	✅
Yolov8	n	segmentation	✅
Yolov8	n	pose	✅
Yolo11	n	detection	✅
Yolo11	n	segmentation	✅
Yolo11	n	pose	✅
Yolo11	n	obb-detection	✅

Important

Before running each model, you need to execute the serialization program to export the .engine file.

Preparation

FOR Jetson Orin Series

1. install source

sudo bash -c 'echo "deb https://repo.download.nvidia.com/jetson/common r36.4 main" >> /etc/apt/sources.list.d/nvidia-l4t-apt-source.list'
sudo bash -c 'echo "deb https://repo.download.nvidia.com/jetson/t234 r36.4 main" >> /etc/apt/sources.list.d/nvidia-l4t-apt-source.list'

2. install dependencies

sudo apt update
sudo apt install tensorrt cudnn 

3. clone package

cd your_workspace/src/
git clone https://github.com/vanstrong12138/trt_yolo.git
cd ..
colcon build

FOR Desktop PC/Laptop

TODO: setup method

Run

1. source workspace

source your_workspace/install/setup.bash

2. run yolov8n detection example

• first, for first run, you need to run the model serialization program to export the .engine file

ros2 run trt_yolo yolov8n_det.launch.py mode:=serialize wts_path:=/path/to/yolov8n_det.wts
# this step will generate the yolov8n_det.engine file in the install directory
# it will take about 10 minutes

• second, when engine file is generated, run the detection node

ros2 run trt_yolo yolov8n_det.launch.py labels_file:=/path/to/coco.txt image_topic:=/your/color/image/topic

3. run yolo11n segmentation example

• first, for first run, you need to run the model serialization program to export the .engine file

ros2 run trt_yolo yolo11n_seg.launch.py mode:=serialize wts_path:=/path/to/yolo11n_seg.wts
# this step will generate the yolo11n_seg.engine file in the install directory
# it will take about 10 minutes

• second, when engine file is generated, run the detection node

ros2 run trt_yolo yolo11n_seg.launch.py labels_file:=/path/to/coco.txt image_topic:=/your/color/image/topic

4. for advance usage, you can run the following command to see more options

# --show-args for detail options
ros2 run trt_yolo yolo11n_seg.launch.py --show-args

Acknowledgements

• Yolov8
• Yolo11
• tensorrtx