Using Computer Vision to Develop a Robotic Arm Poker Dealer

A team from the University of Manchester has developed a robotic poker dealer using computer vision technologies like YOLOv8 and Roboflow to manage the game independently and accurately. The project aims to eliminate human error and potential bias in poker by creating a robotic arm with six degrees of freedom, capable of dealing cards, tracking chips, and recognizing player moves in real time. The system utilizes a dual-camera setup with Android phones to provide different views of the poker table, enabling detailed analysis of game elements such as cards and chips. Using custom-trained models for object detection and instance segmentation, the team ensures accurate identification and action-taking based on the game state, with a multi-threaded architecture facilitating real-time processing. The project not only focuses on the technical integration of vision and motion systems but also plans future enhancements to enable the robotic arm to participate as a player, enhancing its capabilities beyond mere dealing.