The user is given a choice between predicting results for a single driver or for all participants in a session. Shown below are the menus for both and example inputs.
F1_clip_1.mp4
F1_clip_2.mp4
The entire frontend UI is built using React. GET and POST requests are made to the backend server which runs a custom Python API using the trained model. The model used was an XGBoost Regressor from the open-source xgboost library in Python. I trained an instance of this model on a custom dataset which includes data from the past decade of the sport. The relevant data was obtained programatically from theOehrly's fastf1 Python library (https://github.com/theOehrly/Fast-F1).