YOLOX-ROS

YOLOX + ROS2 Foxy demo

🔼 Unity + YOLOX-ROS Demo

Supported List

Base	ROS2 C++	ROS2 Python
PyTorch		✅
PyTorch(CUDA)		✅
PyTorch(CUDA-FP16)		✅
TensorRT (CUDA)	✅
OpenVINO	✅	✅
ONNX Runtime		✅
TFLite		✅

Installation & Demo

Python (PyTorch)

Requirements

• ROS2 Foxy
• OpenCV 4
• Python 3.8 (Ubuntu 20.04 Default)
• PyTorch >= v1.7
• YOLOX v0.3.0
• bbox_ex_msgs

Installation

Install the dependent packages based on all tutorials.

STEP 1 : Download from GitHub

mkdir -p ~/ros2_ws/src
cd ~/ros2_ws/src
git clone https://github.com/Ar-Ray-code/yolox_ros.git --recursive

STEP 2 : YOLOX Installation (yolox_ros_py)

pip3 install yolox

STEP 3 : Install YOLOX-ROS

source /opt/ros/foxy/setup.bash
sudo apt install ros-foxy-v4l2-camera
# source /opt/intel/openvino_2021/bin/setupvars.sh # <- Using OpenVINO
colcon build --symlink-install # weights (YOLOX-Nano) files will be installed automatically.

Automatic download weights

• yolox_nano.onnx by Megvii-BaseDetection/YOLOX
• yolox_nano.pth by Megvii-BaseDetection/YOLOX
• model.onnx by Kazuhito00
• model.tflite by Kazuhito00

(Step 3) Using CUDA

If you have NVIDIA Graphics, you can run YOLOX-ROS on GPU.

Additional installing lists

• NVIDIA Graphics Driver
• CUDA toolkit (11.0)
• torch+cuda

source /opt/ros/foxy/setup.bash
sudo apt install ros-foxy-v4l2-camera
colcon build --symlink-install # weights (YOLOX-Nano) files will be installed automatically.

Step : Demo

Connect your web camera.

source /opt/ros/foxy/setup.bash
source ~/ros2_ws/install/local_setup.bash
ros2 launch yolox_ros_py yolox_nano_cpu.launch.py # <- CPU (PyTorch)
# ros2 launch yolox_ros_py yolox_nano.launch.py # <- GPU (PyTorch)
# ros2 launch yolox_ros_py yolox_nano_onnx.launch.py # <- ONNXRuntime

# OpenVINO -------------------------------------
# source /opt/intel/openvino_2021/bin/setupvars.sh
# ros2 launch yolox_ros_py yolox_nano_openvino.launch.py

C++

Check this URL.

Topic

Subscribe

• image_raw (sensor_msgs/Image)

Publish

• bounding_boxes: Output BoundingBoxes like darknet_ros_msgs (bboxes_ex_msgs/BoundingBoxes)

  ※ If you want to use darknet_ros_msgs , replace bboxes_ex_msgs with darknet_ros_msgs.

Parameters

• Check launch files.

Composition

• Supports C++ only.

YOLOX-ROS + ?

Examples

Web Camera (v4l2-camera)

• yolox_nano_onnx.launch.py

ros2 launch yolox_ros_py yolox_nano_onnx.launch.py video_device:=/dev/video0

Unity

• Ar-Ray-code/RenderTexture2ROS2Image

Gazebo

• yolox_nano_onnx_gazebo.launch.py

ros2 launch yolox_ros_py yolox_nano_onnx_gazebo.launch.py

YouTube

• yolox_nano_onnx_youtube.launch.py
• Ar-Ray-code/YouTube-publisher-ROS2

# git clone https://github.com/Ar-Ray-code/YOLOX-ROS.git --recursive
vcs import . < YOLOX-ROS/youtube-publisher.repos
pip3 install -r YOLOX-ROS/requirements.txt
pip3 install -r YouTube-publisher-ROS2/requirements.txt
cd ..
colcon build --symlink-install --pacakges-select yolox_ros_py bboxes_ex_msgs youtube_publisher
source install/setup.bash

# run launch.py
ros2 launch yolox_ros_py yolox_nano_onnx_youtube.launch.py

Reference

• YOLOX (GitHub)

@article{yolox2021,
  title={YOLOX: Exceeding YOLO Series in 2021},
  author={Ge, Zheng and Liu, Songtao and Wang, Feng and Li, Zeming and Sun, Jian},
  journal={arXiv preprint arXiv:2107.08430},
  year={2021}
}

Contributors

• Ar-Ray
• fateshelled
• Kazuhito00

About writer

• Ar-Ray : Japanese student.
• Blog (Japanese) : https://ar-ray.hatenablog.com/
• Twitter : https://twitter.com/Ray255Ar