A Message to our Community about COVID-19 >>
Sign In
5
Special Offer
FaceMe® The World’s Top Cross-Platform AI Facial Recognition Engine

FaceMe® Accuracy from NIST

National Institute of Standards and Technology (NIST)

NIST FRVT 1:1
VISA True Acceptance Rate
(@FAR 1E-4)
99.70%
VISA True Acceptance Rate
(@FAR 1E-6)
98.95%
VISA Border True Acceptance Rate
(@FAR: 1E-4)
99.06%
WILD True Acceptance Rate
(@FAR 1E-4)
97.17%

* CyberLink FaceMe® is ranked 18th globally and ranked 8th in NIST FRVT 1:1(WILD 1E-4) on Mar 2020. if China & Russia vendors are excluded.

* VISA images are photos taken from the passport. The mean interocular distance (IOD) in the VISA images are 69 pixels.

* Wild images are non-constraint, photojournalism-style photos. Resolution varies widely. A wide yaw and pitch angle may be present.

FaceMe® Accuracy

FaceMe® Model
Ultra High Precision
(UH3)
Very High Precision
(VH)
High Precision Model
(H3)
Recommended Platform
Server, Workstation
PC, Premium Mobile
Mobile, Light-weight IoT/AIoT Devices
True Acceptance Rate
(@FAR 1E-4)
99.05%
98.95%
98.35%
True Acceptance Rate
(@FAR 1E-5)
99.05%
98.95%
97.03%
True Acceptance Rate
(@FAR 1E-6)
98.89%
97.99%
94.98%
Model Size (MB)
250 MB
17 MB
6.7 MB
Template Size (KB)
3 KB
3 KB
3 KB

Performance Numbers

For Workstation or Server

CPU
Core i7-7700K
Core i7-7700K
GPU / VPU
NV RTX 2080 Ti
FaceMe® Extraction Model
UH3
VH
H3
UH3
VH
H3
Execution Time (ms)
Detection + Extraction
69.5 ms
15.9 ms
14.4 ms
19.3 ms
13.5 ms
12.8 ms
Frames per second (fps)
14.4 fps
62.7 fps
69.6 fps
51.7 fps
74.0 fps
78.3 fps
Face Match Time (ms)
0.0024 ms
0.0024 ms
0.0024 ms
0.0024 ms
0.0024 ms
0.0024 ms
CPU
Core i7-7700K
GPU / VPU
FaceMe® Extraction Model
UH3
VH
H3
Execution Time (ms)
Detection + Extraction
69.5 ms
15.9 ms
14.4 ms
Frames per second (fps)
14.4 fps
62.7 fps
69.6 fps
Face Match Time (ms)
0.0024 ms
0.0024 ms
0.0024 ms
CPU
Core i7-7700K
GPU / VPU
NV RTX 2080 Ti
FaceMe® Extraction Model
UH3
VH
H3
Execution Time (ms)
Detection + Extraction
19.3 ms
13.5 ms
12.8 ms
Frames per second (fps)
51.7 fps
74.0 fps
78.3 fps
Face Match Time (ms)
0.0024 ms
0.0024 ms
0.0024 ms

For Industrial PC

CPU
Celeron G4920
Celeron G4920
Jetson Nano
Jetson TX2
GPU / VPU
1x VPU (Movidius Myriad X)
8x VPU (Movidius Myriad X)
FaceMe® Extraction Model
UH3
VH
H3
UH3
VH
H3
VH
UH3
Execution Time (ms)
Detection + Extraction
100.5 ms
54.4 ms
46.4 ms
26.2 ms
19.8 ms
18.8 ms
67.4 ms
92.3 ms
Frames per second (fps)
10.0 fps
18.4 fps
21.5 fps
38.2 fps
50.6 fps
53.2 fps
14.8 fps
10.8 fps
Face Match Time (ms)
0.0036 ms
0.0036 ms
0.0036 ms
0.0036 ms
0.0036 ms
0.0036 ms
CPU
Celeron G4920
GPU / VPU
1x VPU (Movidius Myriad X)
FaceMe® Extraction Model
UH3
VH
H3
Execution Time (ms)
Detection + Extraction
100.5 ms
54.4 ms
46.4 ms
Frames per second (fps)
10.0 fps
18.4 fps
21.5 fps
Face Match Time (ms)
0.0036 ms
0.0036 ms
0.0036 ms
CPU
Celeron G4920
GPU / VPU
8x VPU (Movidius Myriad X)
FaceMe® Extraction Model
UH3
VH
H3
Execution Time (ms)
Detection + Extraction
26.2 ms
19.8 ms
18.8 ms
Frames per second (fps)
38.2 fps
50.6 fps
53.2 fps
Face Match Time (ms)
0.0036 ms
0.0036 ms
0.0036 ms
CPU
Jetson Nano
GPU / VPU
FaceMe® Extraction Model
VH
Execution Time (ms)
Detection + Extraction
67.4 ms
Frames per second (fps)
14.8 fps
Face Match Time (ms)
CPU
Jetson TX2
GPU / VPU
FaceMe® Extraction Model
UH3
Execution Time (ms)
Detection + Extraction
92.3 ms
Frames per second (fps)
10.8 fps
Face Match Time (ms)

* 1080p images, 1 face per image

* Face Match Time: Time taken to match two templates. A template extracted face image can enable a person to be identified from a large database. For example, matching a template with 1,000 people would require 1,000 separate matches.

For Mobile Devices & IoT/ AIoT Devices

SoC
Snapdragon 845
Snapdragon 650
Snapdragon 410
RK3399
A12X
Device
Google Pixel 3
Sony Xperia X
iPad Pro 12.9” 2018
Extraction Model
VH
H3
VH
H3
VH
H3
VH
H3
VH
Execution Time (ms)
Detection + Extraction
41.9 ms
27.8 ms
163.7 ms
112.3 ms
343.7 ms
175.3 ms
41.9 ms
27.8 ms
163.7 ms
Frames per second (fps)
23.9 fps
36.0 fps
6.1 fps
8.9 fps
2.9 fps
5.7 fps
23.9 fps
36.0 fps
6.1 fps
Face Match Time (ms)
0.0009 ms
0.0009 ms
0.0075 ms
0.0065 ms
0.0080 ms
0.0080 ms
0.0009 ms
0.0009 ms
0.0075 ms
SoC
Snapdragon 845
Device
Google Pixel 3
Extraction Model
VH
H3
Execution Time (ms)
Detection + Extraction
41.9 ms
27.8 ms
Frames per second (fps)
23.9 fps
36.0 fps
Face Match Time (ms)
0.0009 ms
0.0009 ms
SoC
Snapdragon 650
Device
Sony Xperia X
Extraction Model
VH
H3
Execution Time (ms)
Detection + Extraction
163.7 ms
112.3 ms
Frames per second (fps)
6.1 fps
8.9 fps
Face Match Time (ms)
0.0075 ms
0.0065 ms
SoC
Snapdragon 410
Device
Extraction Model
VH
H3
Execution Time (ms)
Detection + Extraction
343.7 ms
175.3 ms
Frames per second (fps)
2.9 fps
5.7 fps
Face Match Time (ms)
0.0080 ms
0.0080 ms
SoC
RK3399
Device
Extraction Model
VH
Execution Time (ms)
Detection + Extraction
41.9 ms
Frames per second (fps)
23.9 fps
Face Match Time (ms)
0.0009 ms
SoC
A12X
Device
iPad Pro 12.9” 2018
Extraction Model
H3
VH
Execution Time (ms)
Detection + Extraction
27.8 ms
163.7 ms
Frames per second (fps)
36.0 fps
6.1 fps
Face Match Time (ms)
0.0009 ms
0.0075 ms

* 720p images, 1 face per image

FaceMe® Technology

Features List
Face Detection
Yes
Face Recognition
Yes
Face Tracking
Yes
Face Attributes Recognition
Gender, Age, Emotion, Pose
Real-time Video Support
Yes
RTSP/H.264 Support
Yes
Emotion Detection
Happy, Sad, Angry, Surprised, Neutral
Images Pre-Processing
Lighting Enhancement, Edge Sharpening, Upsampling
Anti-Spoofing with 3D-Cameras
Yes
Anti-Spoofing with 2D-Cameras
(with user interactions)
Yes
Anti-Spoofing with 2D-Cameras
(without user interactions)
Yes
Chipsets (CPU/GPU/VPU)
Intel®64, ARM64, NVIDIA® GPU, NVIDIA® Jetson, Intel®Movidius VPU
Programming Interface
C++, HTTP, C#, Perl
AI Inference Engines
TensorFlow, NCNN, OpenVINO, CoreML

Images/ Cameras Conditions

Minimum
Recommended
Face Pose – Yaw
< 60°
< 45°
Face Pose – Pitch
< 50°
< 30°
Face Pose – Roll
< 60°
< 45°
Lighting
450 lux
550 lux
Face Size
(pixels between eyes)
36 pixels
50 pixels
Face Enrollment
1 image
5 images - Straight, Left/Right (15°~45°), Up (5°~30°), Down (15°~30°)

Anti-spoofing using 3D-Depth Camera

Supported 3D-Depth Cameras
Intel® RealSense, Orbbec® Astra Pro, Himax®, Altek®, eYs3D®
True Positive Rate
98.2%
True Negative Rate
100%

Fast Search Algorithm for an Enormous Database

People in Database
Time to Search in Database (ms)
One-by-one Compare
Fast Search
1 Million People
1,860 ms
1.2 ms
6 Million People
11,200 ms
1.6 ms

* With Fast Search algorithm, FaceMe can search a face within a 6 million person database for about 1.6 ms.

* Tested on a i7-7700K with 24GB memory PC. A database of 30 million faces would require 128GB memory.

Accuracy of Face Attributes Detection

Detection Attributes
Accuracy Rate
Gender
98%
Emotion
Up to 86%
Age
5.8 y/o (Mean Average Error)

* Tested with 68,000 images.

Have questions or need to know more about CyberLink technology solutions?

Contact Us