Magic Leap SuperPoint Pretrained Network

The demo implements a sparse optical flow point tracker. It uses three main components:

1. SuperPointFrontend: A helper class that runs the SuperPoint network on images to compute points and descriptors.
2. PointTracker: A class that forms tracks by finding sequential pair-wise nearest neighbors using two-way matching of descriptors.
3. VideoStreamer: A helper class that handles input from three stream types: a directory of images (e.g., .png, .jpg), a video file (e.g., .mp4, .avi), or a USB Webcam.

The output is a sparse optical flow visualization.

To run the SuperPoint demo, you need Python >= 3.4, OpenCV, and PyTorch >= 0.4. You can install the required libraries using the following commands:

pip install opencv-python
pip install torch