C++ Implementation of SiamMask
You can use the models (with the refine module) trained with the original repository foolwood/SiamMask for inference in C++. Just Follow the instruction in jiwoong-choi/SiamMask to convert your own models to Torch script files.
Or you can download pretrained Torch scripts. These files are converted from the pretrained models (SiamMask_DAVIS.pth and SiamMask_VOT.pth) in the original repository.
git clone --recurse-submodules https://github.com/nearthlab/SiamMaskCpp
cd SiamMaskCpp
mkdir models
cd models
wget https://github.com/nearthlab/SiamMaskCpp/releases/download/v1.0/SiamMask_DAVIS.tar.gz
wget https://github.com/nearthlab/SiamMaskCpp/releases/download/v1.0/SiamMask_VOT.tar.gz
tar -xvzf SiamMask_DAVIS.tar.gz
tar -xvzf SiamMask_VOT.tar.gz
Before building demo, make sure the following command prints out the correct path to torch install directory.
python3 -c "import torch; print(torch.__path__[0])"
# /path/to/lib/python3.x/site-packages/torch
cd SiamMaskCpp
mkdir build
cd build
# specify -DTORCH_PATH=/path/to/lib/python3.x/site-packages/torch if cmake fails to detect PyTorch automatically
cmake ..
make
# Move the executable file to the repository directory
mv demo ..
cd SiamMaskCpp
./demo -c config_davis.json -m models/SiamMask_DAVIS tennis
./demo -c config_vot.json -m models/SiamMask_VOT tennis