Live Detection System

Face recognition access monitoring, built from the camera up.

Neo watches a camera feed in real time, checks every face against a roster of people who are allowed to be there, and flags anyone it doesn't recognize.

Real annotated output from the detection pipeline: facial landmark meshes and UNKNOWN labels drawn over a crowd of unenrolled faces
Real output from detect_faces() — the same pipeline the live dashboard runs. See more on the demo page.

See what the system sees.

01 Capture

A camera watches an entryway, correcting for real-world lighting.

02 Understand

Every face is compared against a roster of enrolled people, GPU-accelerated.

03 Respond

Anyone it doesn't recognize is flagged on a live dashboard, in real time.

See the full detection pipeline →

Built to be honest about what it is.

Deployment

Two cameras, one machine, no cloud infrastructure — running on real hardware right now, not a demo environment.

Authentication

Passwords are bcrypt-hashed. Two-factor authentication (TOTP + recovery codes) is implemented for every account type — production enforcement is currently controlled by a runtime flag, not always on.

Encryption

Every connection uses HTTPS (TLS). Enrolled face data is encrypted at rest with its own key.

Read the full security architecture →

Who's behind this

Who's behind Neo Innovations

Neo Innovations is built and run by one person — a student working toward a career in cybersecurity, building, testing, and documenting every part of this system end to end. There's no company behind it yet: no legal entity, no funding, no customers. It's a real, working system at an early stage, not a product with a sales team behind it.

Read the full story

Want to see it in action, or talk about the project?

See Demo Get in Touch