Build solutions with embedded video analytics
using FindFace SDK

FindFace SDK is a C library. It’s designed to enable developers to add video analytics functionality to their products and services. Without spending resources on developing your own technology.

ADVANTAGES

Top performer in NIST benchmarks (May 2021). Listed in Top-3 in IARPA FRPC, etc.
Feature vector extraction up to ~25 faces per second on CPU, and ~240 faces per second on GPU. Search/verification speed is13.4 mln faces per second.
A well-documented library allows you to start writing code right away. FindFace SDK can be easily integrated into your development process.
Your software can identify and verify faces with or without being connected to the internet.
Liveness check eliminates possible fraud attempts using printed or on-screen images.
Attributes for faces: age, gender, emotions, beard, glasses, etc. For silhouettes: clothes type and color (upper and lower), headwear. For cars: color, make and model, body type, plate.
Several neural networks are available for face recognition. For example, depending on the task, you can choose a faster, but less accurate neural network, or vice versa.
Phenomenal data processing speed.
Approximately 10 times faster than CPU.

LIVENESS

‘Liveness’ check establishes whether the detected face belongs to a live person or whether it’s an image. NtechLab’s Liveness technology uses a passive detection method, which is perfect for access control tasks, and authorization via mobile applications.
Fake face 2%
Fake face 98%
Anti-spoofing system that allows you to distinguish a living person from an image in FindFace SecurityРамкаРамка
Advantages

Does not require actions from visitors (smile, blink, or other interactions with the system).

Works with existing equipment (no need to install specialized devices such as 3D cameras, thermal imagers and others).

Icon
Icon
LOW HARDWARE REQUIREMENTS

Attributes

FindFace SDK allows you to extract object attributes from video and images for advanced analytics and for searching objects simply by specifying parameters.
Age: 28 Gender: male Beard: 98% Glasses: 78%
Find faces by feature set in FindFace SecurityРамка
IconAttributes for faces
IconAge
IconBeard
IconMask
IconHead turn
IconBrightness
IconGender
IconGlasses
IconEmotions
IconImage quality
IconSharpness
IconAttributes for silhouettes
IconClothes type, upper
IconClothes color, upper
IconHeadwear
IconClothes type, lower
IconClothes color, lower
IconAttributes for cars
IconBody type
IconColor
IconCountry of the plate
IconModel, make
IconPlate number
IconRegion of the plate

TYPICAL CASES

E-GATE
E-GATE - automatic border control system using FindFace technology
AUTHENTICATION
Face recognition SDK during authentication
ACCESS CONTROL
Access Control with FindFace SDK
WEARABLES
Face Recognition SDK on Wearable Devices
GET FINDFACE SDK TRIAL VERSION
Fill out the form and get 14-day free access to the library

DETECTOR OPERATION EXAMPLES

Detecting a face means locating its bounds within a digital image. Even if a face is semi-covered, blurred or partially turned away from the camera, FindFace detector will identify it. Our intelligent software outperforms the competition illustrated by outstanding results on low-resolution sources and in limited visibility conditions.

FACE DETECTION
AT A DISTANCE
Distant face detection with FindFace SDK
SEMI-COVERED FACE
DETECTION
Detecting half-covered faces with FindFace SDK
IDENTIFICATION AT AN ANGLE
Detecting faces turned away from the camera using the FindFace SDK
FACE DETECTION
UNDER COMPLEX CONDITIONS (POOR LIGHTING, BLUR)
Detecting faces in difficult conditions of poor lighting and blur

Feature vector extraction and verification

Based on the image of an object (a face, a silhouette, or a car), the neural network builds a feature vector, a descriptor, which is an array of numbers. It is impossible to restore the original image of the object from the feature vector, but the algorithm establishes the similarity between the two images based on the proximity of the vectors.

Look at real photos of verification and evaluate the error-free operation of the algorithm in difficult conditions, with various factors that change appearance that occur in real life.

FACE FROM CCTV CAMERA

KNOWN FACE

FACE FROM CCTV CAMERA

KNOWN FACE

FACE FROM CCTV CAMERA

KNOWN FACE

Half-covered face
Recognizing people whose faces are covered using FindFace SDKFindFace Face Recognition on Closed-Face Images
Match: 83%
Blurred image
Recognize people in motion blur with FindFace SDKFindFace Face Recognition on Blurry Images
Match: 77%
Half-face
FindFace Half Face Recognition from Surveillance CamerasFace recognition of half-faced people using the FindFace SDK
Match: 74%
Sharp head tilt
Face recognition with strong head tilt using FindFace SDKFace Recognition with Strong Head Tilt with FindFace SDK
Match: 83%
Sunglasses
Face recognition in sunglasses with FindFace SDKRecognizing people with glasses using the FindFace SDK
Match: 77%
Poor lighting
Face Detection in Low Light with FindFace SDKRecognizing half-faced people with the FindFace SDK
Match: 77%

CODE SAMPLE

Written in C, the face recognition SDK is easy to use. The package includes samples in C, CMake file and documentation.

GET FINDFACE SDK
TRIAL VERSION
Fill out the form and get 14-day free access to the library.
REQUEST TRIAL
TECHNICAL
SUPPORT
Should you have any questions on the FindFace SDK installation and usage, improvement suggestions or the need for customization, contact our support team by [email protected]