Modulai built a solution for Passion Football's ICONS app that makes performance analytics available to players at every level, not just professionals. It lets any football player measure their shot speed using nothing but a smartphone camera, combining object detection, object tracking, and trigonometry to translate the ball's movement on screen into real-world speed.
Challenge
The ICONS app by Passion Football lets football players at every level develop their skills and challenge others. Users work through a range of exercises that measure essential football skills, designed to be fun and interactive while delivering analytical insights and suggesting drills that target weak spots. One of these lets players measure and improve their shooting power by recording a video of their kick.
Solution
An object detector locates the bounding boxes of all footballs and goals in the video. An object tracker then refines these detections to reflect real-world constraints, such as the fact that a football follows a predictable trajectory, by applying a constant velocity model and a 2D nearest-neighbor filter. Finally, the known size of the football is used together with trigonometry to translate the ball's speed in the video into real-world speed. The app communicates over a REST API with a cloud service that runs inference and handles all the necessary calculations.
Tools
The codebase was built in Python, with the option to export the model to mobile via ONNX. DVC handled data and model version control, and the object detection data was annotated using a SaaS tool. PyTorch and PyTorch Lightning were used to train the object detection model with data augmentation. The object tracker uses Kalman filters and was developed in house.
Value created
ICONS is in production and live in App Store, helping players at every level develop their game. The wider aim is to get more kids moving and cut down on sedentary time.
Learn more
