I have an MP4 recording that shows two bikes riding in the same frame. From that single video I need to extract, with good accuracy, both • the gap in time between the bikes (resolution: 0.01 s) • the physical gap in distance (resolution: centimetres). There are no coloured bibs or markers to help, so the solution has to rely on computer-vision detection and tracking alone (OpenCV, YOLO, DeepSORT, or any comparable model you prefer). A simple calibration step is fine—e.g. I can click on a reference object of known length in the first frame so the algorithm converts pixel distance to real-world centimetres. What I expect to receive • A runnable script or lightweight app (Python, C++, or similar) that lets me load the MP4, define the scale once, optionally draw or confirm the initial bounding boxes around each bike, and then produces a CSV (or JSON) with timestamp, inter-bike distance, and inter-bike time gap for every analysed frame. • A brief read-me covering required libraries, how to install them, and how to operate the tool. • The ability to process videos of at least 1080p at normal cycling speeds without manual frame-by-frame correction after the initial setup. I am happy to test early prototypes and give feedback, so iterative delivery is welcome as long as the final tool satisfies the precision above and runs on Windows 10.