Contact person: Tapo Cash-Handling Event Detection

Phone:Show

Email:Show

Location: Vallensbæk Strand, Denmark

Budget: Recommended by industry experts

Time to start: As soon as possible

Project description:
"I run several restaurants and activity-based venues that rely on TP-Link Tapo cameras positioned at each till. I need a software layer that continuously watches those live RTSP feeds, recognises moments when cash is handled, and highlights anything that looks suspicious—missing hand-offs, unusually long counts, cash leaving the frame, and similar behaviours.

The system has to work in real time while also analysing the recorded footage later, storing every detected event with its timestamp so we can audit a full shift whenever necessary. Whenever an event is logged it should appear instantly on a web-based dashboard; no e-mails or mobile push for now, just a clear alert list with clip thumbnails, camera name, and quick playback.

You’re free to build with tools you know best—OpenCV, YOLOv8, TensorFlow, PyTorch, or similar—so long as the final solution can run on a small Windows or Linux box already on-site and connect to multiple Tapo streams at once. Low latency, solid accuracy, and minimal false positives are the priorities.

Deliverables
• Trained cash-handling / suspicious-activity model and all source code
• Possibility to handle other custom events in the future
• Lightweight dashboard UI that lists, filters, and replays events
• Installation script and step-by-step setup notes for a fresh machine
• Short test session proving detection works on both live and archived video

The project is complete when the dashboard reliably flags genuine cash-handling events from at least three Tapo cameras simultaneously and allows us to review those clips without manual searching." (client-provided description)