README.md

ROS 2 Real-Time Image Segmentation with ONNX Runtime

Overview

This ROS 2 package performs real-time image segmentation using the ONNX Runtime. Designed for robotics applications, it processes camera input and generates segmented output in real-time.

---

Features

• Real-time image segmentation using ONNX Runtime.
• Compatible with ROS 2 (tested with Humble).
• Supports Yolo ONNX segmentation model.
• Configurable parameters for input, output topics, and model path.

---

Requirements

• ROS 2 Humble
• ONNX Runtime (Auto Installed)
• OpenCV 4.5 or later

Usage

1. Clone the Repository into ROS 2 Workspace Navigate to your ROS 2 workspace's src directory and clone the repository:

git clone https://github.com/Leeseunghun03/ros_onnx_segmentation.git

Build the workspace:

colcon build --symlink-install

2. Customize Parameters Adjust parameters such as class, input_topic, output_topic, and other settings in the provided configuration file or via command-line arguments.
3. Run the Camera Package Start your camera package to publish raw image data to a topic (e.g., /camera/image_raw).
4. Launch the Segmentation Package Run the segmentation package using the launch file:

ros2 launch ros_onnx_segmentation segmentation_launch.py

Example (Custom model using Yolov8n-seg / without GPU)

Reference

• https://github.com/winxos/yolov8_segment_onnx_in_cpp.git

• https://github.com/ms-iot/ros_msft_onnx.git