Using Computer Vision to Play Dreidel

Salo Levy, a new employee at Roboflow, combined his enthusiasm for the Jewish holiday Chanukah with a computer vision project to create a model that recognizes the Hebrew characters on a Dreidel. The game Dreidel involves spinning a top marked with Hebrew letters, each dictating a player's action based on the side that lands face up. Levy's project aimed to help players by using computer vision to identify these characters, making it easier to learn the game. He collected and annotated 569 images, ensuring diverse examples for effective model training. Despite challenges with class imbalance, he successfully created a model that can be tested with a webcam. Levy shared his dataset and model on Roboflow Universe, inviting others to build upon his work, potentially creating applications that automatically score the game or track character frequency. This initiative highlights the potential of integrating technology with cultural traditions to enhance learning and engagement.