Key Imports and Technologies

GFPGAN

  • GFPGAN is used for face restoration and enhancement
  • It stands for Generative Facial Prior GAN
  • The system uses it to improve the quality of face images before saving them
  • Pretrained weights such as GFPGANv1.3.pth are required for operation

MongoDB

  • MongoDB serves as the primary database for the application
  • It stores identity information including face embeddings
  • Using a document database allows for flexible storage of person metadata
  • The system connects to MongoDB running on the local machine by default

InsightFace

  • InsightFace provides the models for face detection and recognition
  • It generates embeddings that are used to compare faces in the database
  • The buffalo_l model is used for balanced speed and accuracy
  • It is configured to run on CUDA enabled GPUs for real time performance

ZED SDK

  • The ZED SDK is required to interface with the ZED2i camera
  • It provides depth sensing capabilities that help filter out distant faces
  • The SDK also offers object detection features used in later iterations