DEVELOP PRODUCTS
WITH FINDFACE SDK

FUNCTIONALITY

The FindFace SDK is a C library that provides access to the cutting-edge face recognition technology based on neural networks. The SDK allows you to quickly and accurately solve the 3 key tasks of face recognition:

ADVANTAGES

IARPA FRPC winner, in the top among all NIST benchmarks
Biometric sample extraction ~25 faces per second on CPU, or ~240 faces per second on GPU; search/verification ~13.4 mln faces per second. GPU-acceleration support: ~10 times quicker compared to CPU.
Simple integration with your product
Identify and verify faces without being connected to the internet
Liveness detection: protect your system from photo-attacks and spoofing.
Estimate age, emotion, gender, image quality, blur, face angle, presence of a beard and glasses for each face
Two neural networks for your choice: prioritize performance or accuracy.
Phenomenal data processing speed
GPU acceleration support

SYSTEM REQUIREMENTS

CPU/GPU

CPU with AVX support/NVidia video card with 6 Gb of video memory for more accurate calculations, 3 Gb of video memory for quicker calculations


RAM

~3.5 GB RAM


OS

Windows 7, 8, 10 (x64), Linux Ubuntu (x64)


TYPICAL CASES

E-GATE
AUTHENTICATION
ACCESS CONTROL
WEARABLES
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 it in a digital image. Even though a face is small, semi closed, blurred or turned away from the camera, a good detector will find it.

Detection examples from NtechLab:
FACE DETECTION
AT A DISTANCE
SEMI CLOSED FACE
DETECTION
LARGE ANGLES
FACE DETECTION
UNDER COMPLEX CONDITIONS (POOR LIGHTING, BLUR)

KEY FEATURES

Highest accuracy even with the factors that change appearance:

0
POOR LIGHTING
1
COVERED HEAD (hat, scarf, shawl)
2
GLASSES, BEARD, MUSTACHE
3
SEMI CLOSED FACE
4
AGE (photo taken 5 years ago)
5
VARIABILITY OF FACE EXPRESSIONS (smiling/gloomy)
6
DIVERSE FACE ANGLES
7
DIFFERENT CONDITIONS (sober/drunk)
8
DIFFERENT SIZE IN PIXELS

Biometric sample extraction and verification

Biometric sample extraction is a process of obtaining the distinctive features of a particular face from an image using neural networks. The result of this process is a unique array of numbers that describes a face, known as biometric sample, feature vector or face descriptor. It can be further compared to another array of numbers to determine the degree of similarity between the two faces. Look at the real verification cases and acknowledge the algorithm error-free operation in non-ideal conditions with various appearance changes.

FACE FROM SURVEILLANCE CAMERA

KNOWN FACE

FACE FROM SURVEILLANCE CAMERA

KNOWN FACE

FACE FROM SURVEILLANCE CAMERA

KNOWN FACE

Half-covered face
0.jpg0.jpg
Match: 83%
Blurred image
1.jpg1.jpg
Match: 77%
Half-face
2.jpg2.jpg
Match: 74%
Sharp head tilt
3.jpg3.jpg
Match: 83%
Half-face
4.jpg4.jpg
Match: 74%
Sun glasses
5.jpg5.jpg
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.

LICENSING
FindFace SDK is protected by the Sentinel LDK licensing system. FindFace SDK requires purchasing a license for each server it runs on.
request license
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 support@ntechlab.com