School capstone, computer science

Face recognition, shown step by step.

This project demonstrates how a camera image becomes a face embedding, how the backend compares it with saved samples, and why a match is accepted or rejected.

Start registration Run identify
  • 01Detect face
  • 02Encode features
  • 03Compare distance
  • 04Return confidence

What we are building

Face Lab is a full-stack facial recognition demo. The frontend is Astro, the backend is FastAPI, and the recognition layer uses dlib through the face_recognition Python library.

Why it matters

The goal is not a mystery button. The interface shows detection count, distance, confidence, and failure states so the model's decision can be explained during a school presentation.

Workflow

From camera frame to answer

  1. Register a person with clear camera captures or uploaded images.
  2. The API finds the most useful face in the image and stores its 128-value embedding.
  3. Identify compares a new image against saved embeddings and returns the closest known person.
  4. Verify compares two images directly and reports whether they appear to be the same person.