Andrew Sun-Zhuang built a $260 AI pitching machine with 94% strike accuracy

PC: Society for Science/ Image AI generated Andrew Sun-Zhuang, a 7th grader at The Evergreen School in Shoreline, Washington, plays baseball and is constantly trying to improve his batting performance. His hit rate is 95 per cent against a batting machine but only 35 per cent in real life. According to the Society for Science, he decided to train an AI to produce a better pitching machine. Using a Raspberry Pi, a camera, servos and motors, he built VisionStrike, a portable AI-powered pitching machine. It pitched with 86 per cent accuracy within a strike zone for front-toss pitching and threw fastball strikes with 94 per cent accuracy. The final prototype cost only $260.How did Andrew build the $260 AI pitching machineAndrew began by building his own pitching machine. He used a microcontroller, a Raspberry Pi, a camera, servos and motors. The Raspberry Pi later ran the AI model that controls the machine. For the AI itself, Andrew trained a model on 200 different pictures of a batter, himself. The model detected six key points on every image: the end of the bat, the wrist, elbow, shoulder, hip and knee. The model used these points to determine where the bat strike zone should be.Once...