SuperGlue C++ Demo

Building

First, generate TorchScript module files of SuperPoint and SuperGlue by JIT-ing the annotated model definitions.

$ python3 ../jit.py

This should output SuperPoint.zip and SuperGlue.zip.

Building the demo project requires libtorch and OpenCV 3+. Follow the instructions in Installing C++ Distributions of PyTorch for libtorch setup.

Create a build directory and configure CMake.

$ mkdir build && cd build
$ cmake .. -DCMAKE_PREFIX_PATH=<libtorch path>
$ make

Usage

$ ./superglue <image0> <image1> <downscaled_width>

This will measure the average FPS over 50 iterations and outputs matches.png with a visualization of the detected keypoints and matches.