AI Camera Piece & Person Counter

I need an AI-driven vision system that can watch a production line in real time, count every manufactured item that passes the lens, and simultaneously detect whether an operator is present at the station. While computer-vision tools are often used for security surveillance, customer behavior analysis, or traffic monitoring, my requirement is much narrower: rock-solid piece counting and clear person-availability status. Here is what I expect at hand-off: • A trained model (e.g., YOLOv8, TensorFlow, or equivalent) fine-tuned on images of our manufactured parts and typical shop-floor clothing/poses. • Edge-ready code that runs on an IP camera feed (OpenCV or GStreamer pipeline acceptable) and returns two streams of data: item count per frame and boolean person-present flag. • A lightweight dashboard or API endpoint so we can view counts live and pull daily logs. • Setup documentation that lets my in-house team redeploy the model if we swap cameras or add a new SKU. Timeline is flexible—there’s no immediate deadline—so quality and robustness matter more than sheer speed. If you have prior work in factory analytics, image segmentation, or people detection, that background will be valuable.