I want to replace paper timesheets with an automatic clock-in / clock-out system that identifies staff by face the moment they pass our main entrance camera. As soon as a recognised face crosses the frame, the system should log the arrival or departure time against that employee’s profile and store the record in an attendance database I can export to CSV or Excel. Architecture is open: I’m comfortable with a cloud, on-prem, or hybrid deployment—whichever you feel offers the best balance of accuracy, speed and privacy given a single entrance camera. No links to my HR or payroll software are needed for now, so a self-contained dashboard or lightweight web portal that lets me view, edit and download daily, weekly and monthly reports is perfect. Key deliverables • Face-enrolment module for adding and updating employee profiles • Real-time detection and match with ≥95 % accuracy on varied lighting at the doorway • Automatic in/out logic that prevents double entries when someone lingers near the camera • Admin dashboard with search, manual override and CSV/XLS export • Installation guide plus brief training so I can manage enrolments myself • Source code or licence details, and clear instructions for future camera or user expansion Feel free to use OpenCV, DeepFace, TensorFlow, AWS Rekognition, or another proven toolset—just explain the choice and any recurring costs. I will supply the RTSP stream from the entrance camera and can provision a small Windows or Linux box on-site if you recommend an on-prem edge device. Once everything is running smoothly and the test group’s times are logging correctly for a full week, I’ll consider the project complete and release final payment.