Computer Vision for Educational Games

An educational game combined a physical board, letter and number pieces, a camera and projected feedback. The software needed to recognise the pieces and their positions so it could calculate the game state and display the right response.

My contribution I helped develop the computer-vision component: an algorithm using a machine-learning model to locate and identify game pieces from camera images. The recognition output gave the game logic the information it needed to respond to play.

Engineering priorities The component had to support responsive play without excessive computing resources, while leaving room for additional symbols and languages. This required attention to both the recognition method and how its output fitted into the wider game.

Delivered capability The educational game entered production with our recognition work contributing to the system. The component also provided a basis for extending the platform with more symbol sets and games.