VisionLabs - Face Engine



VisionLabs provides face detection and face recognition functions in images and live video streams.

The latest face recognition algorithm of VisionLabs was ranked No. 1 at both 1:1 and 1:N verifications in the NIST FRVT Ongoing benchmark. We have applied the Deep Neural Network (DNN) architecture and FLOWer DNN inference framework, both of which we have developed, to some of our products and solutions. In addition, PyTorch, a deep learning framework, is being developed in only 15 regions worldwide, and VisionLabs is one of them.

VisionLabs provides the following features in collaboration with LG AI Platform:

Vision:Labs Features
Face detection Extraction of facial feature point Face recognition Analysis of facial properties

face engine

face detectionpoint detectionface recognition

analysis of facial properties


Detects the face area within an image and provides information about the detection area.

Extracts 68 facial feature points from the detected face area.

Compares the similarity between the detected face and the face registered in the database to identify users.

Analyzes the properties such as age, gender, emotion, and race in the detected faces.


Engine Structure

The engine provided by VisionLabs is also designed in the same structure as LG AI Platform's existing engine. The VisionLabs engine takes an image as an input and provides an output in JSON format.

The output contains the results for functions set in property settings. You can select properties for the functions you need among the functions shown below.


  • Face detection
    Age, emotion, race, eye conditions (closed or open), head orientation, presence of glasses, age judgment, etc.
  • Face recognition
    You can set a threshold to determine the degree of similarity between face images.


Visionlabs architecture


Examples of Use

VisionLabs' face detection function can be used in various fields such as security and sales.


  • Smart security
    You can integrate the face detection function into the security system to provide the gate access control service through face recognition and the vehicle theft prevention service through driver recognition.
  • Smart store and advertising
    You can use the face property analysis function to analyze data according to gender or age for providing customized advertisements or establishing sales strategies. VisionLabs can recognize visitors together with an Artificial Intelligence (AI) robot and provide the guidance service for those visitors. You can automatically control access to the products that require age verification through the recognition of age and gender.
  • Smart building/home
    VisionLabs can provide access control service together with a smart doorbell/door lock by recognizing family members or pre-registered visitors. You can enhance security by applying VisionLabs to elevators and allowing pre-registered visitors to access only designated floors.