Face recognition

Face Detection is the first half of the face recognition process and consists of three steps: face detection, landmark detection, and face alignment. When an image is input, AI first finds an area of the face, and when a face is detected, it extracts landmark points. The face image is then rotated and edited based on the landmark points to transform it into a form that enables face matching.

Face matching is the remaining half of face recognition after face detection, and consists of feature extraction and feature contrast. After receiving an aligned face image, the AI applies dozens of convolutional layers to the image to transform the face into an N-dimensional vector value. Since this vector value is unique to each person, it is possible to confirm identity by later comparing them with each other.

Face Anti-Spoofing is a function that checks whether the input face image was taken from an actual person who was actually in the location at the time of shooting, and can be operated by all IR, 3D Depth, and RGB cameras. As proven by many applications and industry groups, Alchera's Anti-Spoofing (anti-counterfeiting) technology has a 99% True Positive Rate and is very effective in preventing fraud such as identity theft.

Mask Detection is a function that checks whether the person being photographed is wearing a mask correctly. With an accuracy of 99.70%, Alchera's Mask Detection can even capture so-called “tucks” and “cosks.”

Age, Gender, and Emotion (AGE) is a function that estimates the age, gender, and emotions of the photographed person. Brick-and-mortar stores can implement services such as recommending products by dividing the characteristics of customer groups according to age and gender.