Labeled Faces in the Wild (LFW) is recognized internationally as one of the most challenging databases of face photographs. It contains more than 13,000 images of labeled faces and is used to gauge face recognition accuracy.
Threshold is a score indicating whether faces in two photos are of the same person. Generally, the 0.1% threshold (score with 0.1% FAR) and 0.01% threshold (score with 0.01% FAR) are provided.
Liveness detection can meet the high security requirements of facial recognition in different verification scenarios. It can analyze videos and decide whether they come from a live person or a fake, and guard against presentation attacks such as photos, videos, and static 3D models. Currently, it supports lip movement, motion silent live detection, etc.
Face comparison can calculate the similarity of facial features between two faces and automatically perform identity verification.