I need software that ingests a live camera feed of a roulette wheel, tracks both the wheel and the white ball in real time, and pinpoints the pocket in which the ball finally rests. Because every spin is unique—the croupier’s force varies and the wheel’s speed changes—your solution has to read the motion on-the-fly, update its calculations continuously, and announce the result the instant the ball settles. The core of the job is a real-time tracking system, not an offline simulation. I will be streaming the wheel over video, so the program must overlay or otherwise broadcast its predictions on the same feed to create a seamless live experience. OpenCV, Python, C++, or any comparable computer-vision stack is fine as long as the latency stays low and the detection is reliable across many lighting conditions and wheel designs. Deliverables are simple: a working application (source code included) that accepts a standard video stream, displays the wheel with the running prediction, and outputs the final pocket number once motion stops. A brief setup guide is enough—no elaborate documentation needed, just clear steps to replicate your test environment and run the program on my camera feed.