This project is about understanding the untold story of football games. Raising the stakes by bulding KPIs about players and revealing their performances during live / replays videos.
Please have a look at our project here : kickvision.streamlit.app
The main folder is user_interface. That is all you need to test the project. Dataset : https://www.kaggle.com/datasets/azminetoushikwasi/ucl-202122-uefa-champions-league
- lib/data_processing.py : functions which aim to clean the faces the model was trained on.
- lib/video_processing.py : functions which process the video you pass on the app to predict players and build stats.
- lib/model_83_nik.h5 : model used to predict the players. Shootout to Nikolay for creating it !
- stats/graphics.py : main graphics that are displayed for the predicted players of your video.
- stats/kpi_formulas.py : additional KPIs that you can use in case you want to display extra information on the app.
- stats/stats_preprocessing.py : extra processing steps of the Kaggle UCL dataset.
You can uploaed a < 200 MB video on the app. The model will then predict the faces.
If predicted, you can select multiple players to show their UCL 21/22 statistics.
- Faces recognition
- Faces prediction
- Graphics building
- Clone the repository:
git clone https://github.com/your-username/dynamic-players-insights.git
cd dynamic-players-insights
- Install the required dependencies: pip install -r requirements.txt