Development Iterations

Main

  • This is the initial version of the face recognition system
  • It uses InsightFace for detection and embeddings
  • ZED SDK provides depth information to calculate distance
  • Simple logic is used to decide whether to register a new person based on distance

F2

  • This version introduces several enhancements for stability
  • Face tracking is added to keep track of individuals across frames
  • GFPGAN is integrated to restore and enhance face images
  • A registration buffer ensures that only clear faces are saved to the database
  • Yaw estimation helps in selecting the best face angle for registration

FC3

  • The final iteration focuses on integration and usability
  • A Flask server is included to provide a live video feed
  • Streamlit dashboard connects to this feed for real time monitoring
  • Images are automatically saved to the local disk during registration
  • Logic for matching and creating identities is further refined