Fix camera calibration and player tracking in an existing football tactical-board web app

I am looking for a computer-vision developer to fix and improve an existing football video analysis feature. I already have a working web application and 2D football tactical board. The application processes a football video sequence and transfers the detected player positions onto the 2D tactical board. The basic system is already implemented, but it is currently not accurate and stable enough. Current problems: - The football pitch calibration is inaccurate. - Player positions in the video do not always correspond to the correct positions on the 2D tactical board. - Player tracking is unstable during the video sequence. - Players may be lost, positioned incorrectly, or assigned the wrong tracking identity. - Calibration may become incorrect when the camera moves. I built the application with the assistance of Claude and Codex, but I am not an experienced software developer. I need someone who can inspect the existing code, identify the actual causes, implement the necessary corrections, and integrate the solution into the existing project. I can provide: - a screen recording showing the complete workflow and the current errors - the original football test-video sequence - access to the existing GitHub repository after initial contact - access to the current web application - examples of where the player positions are incorrect Relevant knowledge: - computer vision - football or sports video analytics - camera calibration - homography and perspective transformation - player detection and multi-object tracking - Python and OpenCV - YOLO, Roboflow, ByteTrack, DeepSORT, BoT-SORT, or similar technologies - integration with an existing Next.js web application This is not a request to rebuild the entire application. I want the existing implementation to be analyzed, corrected, and made functional. Please review the attached demonstration before applying and answer: 1. What do you think is causing the calibration and tracking problems? 2. Have you worked on football-field calibration, homography, or player tracking before? 3. What would you inspect or change first? 4. Can you work with and improve an existing codebase? 5. What fixed price would you propose after reviewing the materials? 6. What result would be included in that price? 7. How long would the work take? My initial displayed budget is a placeholder because I do not yet know the actual technical effort. My budget is limited, so please propose the smallest practical solution that makes the existing system work reliably on the supplied test video.