Final project for Computer Vision, CSE 5095/4830. We built a prototype of a system that can track the 3D position of a ping pong ball using a single smartphone camera. We also attempted to track spin by adapting the models and code from https://github.com/cogsys-tuebingen/spindoe.
The src directory contains the code for the main components of our project. It includes our Python modules for object detection, position estimation, kalman filtering, visualization and more.
The notebooks directory contains some of our experiments. It has scripts in .ipynb format used to test on real data and generate visualizations.
The data directory contains data we collected to test our approach. It includes 3 videos with associated configuration files.
To use our code, clone the repository and create a conda environment from the environment.yaml file.
Then, you can run the .ipynb files in our notebooks folder.
The main notebook that runs the object detection, position estimation and kalman filtering over an entire video is in process_video.ipynb. You can change the video by changing the config file used at the beginning of the script.