A Machine Learning and Computer Vision Application to Robustly Extract Winnings from Multiple Lottery Tickets in One Shot

Abstract

Mega Millions and Powerball are among the most popular American lottery games. This article provides a practical software application that can conveniently examine and evaluate several lottery tickets for prizes using just the images. The application accepts as input a directory containing the images of lottery tickets and utilizes machine learning and computer vision to extract lottery ticket data, lottery name, lottery draw date, 5-digit lottery numbers, 2-digit lottery "ball" numbers, and the lottery multiplier. The application also retrieves winning lottery data that corresponds to the lottery draw date using a public database API. This is compared with data collected from each lottery ticket image to establish matches, and the corresponding prize amount is computed. The current version of the application supports GPU usage, and image orientation has no impact on its functionality. It is believed that a considerable portion of the U.S. public participating in the Powerball and Mega Millions lotteries will find such an application beneficial and handy.