The next repository aims to provide a basic c++ script using std 17 over, to do it and consider the speed The code use OpenCv 4.9.0_8 and Onnx 1.17.1 to manipulate the image and inference the model. Note that Opncv don't support a native integration because yolov10 integra A top K layer in their architecture.

1. Download de model you want

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

    python download_model.py  --model {MODEL_SELECTED}

    yolo export model=yolov10n.pt format=onnx

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

1. Using Cmake, Recommended

    mkdir build
    cd build
    cmake ..
    make

2. Run the following command

    ./yolov10_cpp [MODEL_PATH] [IMAGE_PATH]

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

[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 https://arxiv.org/abs/2405.14458