This code is to run a Segment Anything Model 2 ONNX model in c++ code and implemented on the macOS app RectLabel.
sam2_polygon.mp4
Install CUDA, cuDNN, PyTorch, and ONNX Runtime.
Install Segment Anything Model 2, download checkpoints and copy yaml files in sam2/configs/sam2.1 to sam2.
git clone https://github.com/facebookresearch/sam2.git && cd sam2
pip install -e .
cd checkpoints
./download_ckpts.sh
cd ..
cp sam2/configs/sam2.1/*.yaml sam2Install Segment Anything Model 2 CPP Wrapper.
git clone https://github.com/ryouchinsa/sam-cpp-macos.git
cp sam-cpp-macos/export_onnx.py .
cp sam-cpp-macos/david-tomaseti-Vw2HZQ1FGjU-unsplash.jpg .Export an ONNX model and check how the ONNX model works.
python export_onnx.py --mode export
python export_onnx.py --mode import
cp -r checkpoints/sam2.1_tiny sam-cpp-macos
cd sam-cpp-macosDownload exported SAM 2.1 ONNX models.
Build and run.
# macOS
cmake -S . -B build -DONNXRUNTIME_ROOT_DIR=/Users/ryo/Downloads/onnxruntime-osx-universal2-1.20.0
# Ubuntu GPU
cmake -S . -B build -DONNXRUNTIME_ROOT_DIR=/root/onnxruntime-linux-x64-gpu-1.20.0
cmake --build build
# macOS
./build/sam_cpp_test -encoder="sam2.1_tiny/sam2.1_tiny_preprocess.onnx" -decoder="sam2.1_tiny/sam2.1_tiny.onnx" -image="david-tomaseti-Vw2HZQ1FGjU-unsplash.jpg" -device="cpu"
# Ubuntu GPU
./build/sam_cpp_test -encoder="sam2.1_tiny/sam2.1_tiny_preprocess.onnx" -decoder="sam2.1_tiny/sam2.1_tiny.onnx" -image="david-tomaseti-Vw2HZQ1FGjU-unsplash.jpg" -device="cuda:0"