yolov8seg-ort-rs

This project provides a Rust implementation of YOLOv8 segmentation using the ONNX Runtime inference engine. It is based on the YOLOv8 example from the pykeio/ort project.

Requirements

• Rust programming language
• Microsoft ONNX Runtime v1.18.0 or later
• yolov8m-seg pt model

Install the required packages

Run the following commands to install the required packages:

pip install onnxruntime-gpu  # Install GPU version
pip install onnxruntime  # Install CPU version
pip install ultralytics

Prepare your YOLOv8 segmentation model in ONNX format.

1. Download the model

   Visit the Segment Models section in the Ultralytics YOLO documentation to download the desired model. For example, you can download the YOLOv8m-seg model.

2. Convert the model to ONNX format

   Run the following command to convert the YOLOv8m-seg.pt model to onnx format:

   yolo task=segment mode=export model=yolov8m-seg.pt format=onnx

   This command will convert the YOLOv8m-seg model from PyTorch format (.pt) to ONNX format (.onnx).

Download and install the ONNX Runtime

1. Download and install the ONNX Runtime shared library from the official releases page. Make sure to choose the appropriate version for your operating system.

2. Set the LD_LIBRARY_PATH environment variable to the directory containing the ONNX Runtime shared library. For example:

   export LD_LIBRARY_PATH=/path/to/onnxruntime/lib:$LD_LIBRARY_PATH

Installation

Clone this repository:

git clone https://github.com/your-username/yolov8seg-ort-rs.git
cd yolov8seg-ort-rs

Usage

1. Edit src/main.rs

Replace path/to/your/model.onnx with the path to your YOLOv8 in Line:129.

const YOLOV8M_FILE_PATH: &str = "/home/ubuntu/projects/ort/yolov8m-seg.onnx";

And image path in Line:139 with the path to the input image.

let original_img = image::open(Path::new(env!("CARGO_MANIFEST_DIR")).join("data").join("baseball.jpg")).unwrap();

2. Run the YOLOv8 segmentation inference:

./target/release/yolov8seg-ort-rs

Replace path/to/your/model.onnx with the path to your YOLOv8 segmentation model and path/to/your/image.jpg with the path to the input image.

3. The segmentation results will be saved as an image file named output.jpg in the current directory.

Contributing

Contributions are welcome! If you find any issues or have suggestions for improvements, please open an issue or submit a pull request.

License

This project is licensed under the MIT License.

Acknowledgments

• pykeio/ort - Rust wrapper for ONNX Runtime
• Microsoft ONNX Runtime - Open-source inference engine for ONNX models