Yolo V10 cpp ONNX

Notes

Use the yolov10cpp given in the README(https://github.com/THU-MIG/yolov10) ,will get incorrect results as follows:

The cause of the error is that the pre-processing part does not use letterbox, but only uses a simple resize. Therefore, the corresponding post-processing part is also wrong.

After modification, the correct result is as follows:

Prepare the code

1. Download de model you want

• yolov10n
• yolov10s
• yolov10m
• yolov10b
• yolov10l
• yolov10x

    python download_model.py  --model {MODEL_SELECTED}

Install packages

    conda create -n yolov10 python=3.9
    conda activate yolov10

    git clone https://github.com/THU-MIG/yolov10
    cd yolov10

    pip install -r requirements.txt
    pip install -e .

    cd ..

Convert model

    yolo export model=yolov10n.pt format=onnx

Dependencies

1. ffmpeg
2. Opnecv
3. onnxruntime

• MacOs

    brew install ffmpeg 
    brew install opencv
    brew install onnxruntime

• Ubuntu: Unfortunately, onnx runtime is no available using native apt-get

You can use python

sudo apt-get update
sudo apt-get install python3-pip
pip3 install onnxruntime

dotnet

dotnet add package Microsoft.ML.OnnxRuntime

How to run this code

1. Using Cmake, Recommended

    mkdir build
    cd build
    cmake ..
    make

2. Run the following command

static images

    ./yolov10_cpp [MODEL_PATH] [IMAGE_PATH]

realtime

    ./yolov10_cpp_video [MODEL_PATH] [SOURCE]

Future plans

1. Modularize the components.
2. Make a example to video real time.
3. Support Cuda. ?

Reference

[1] Wang, A., Chen, H., Liu, L., Chen, K., Lin, Z., Han, J., & Ding, G. (2024). YOLOv10: Real-Time End-to-End Object Detection. arXiv [Cs.CV]. Retrieved from http://arxiv.org/abs/2405.14458

[2] yolov10cpp Jose Sarmiento https://github.com/DanielSarmiento04/yolov10cpp