Neural Network Models

Here you can see a summary for neural network models created by our Lab and used in FindFace:

Important

The default face biometrics model upon a clean install is jackfruit_480.

Model	Type
face/jackfruit_160	Face biometrics
face/jackfruit_320
face/jackfruit_480
face/ifruit_320
faceattr/age.v1	Age recognition
faceattr/beard.v0	Beard recognition
faceattr/emotions.v1	Emotions recognition
faceattr/gender.v2	Gender recognition
faceattr/glasses3.v0	Glasses recognition
faceattr/medmask3.v2	Face mask detection
faceattr/liveness.alleyn.v2	Liveness standalone service
faceattr/quality.v1	Face quality estimation

Note

The CPU and GPU benchmark setup is the following:

• CPU - Intel® Core™ i7-6700 CPU @ 3.4GHz 4 cores

• GPU - GeForce GTX 1080 Ti

Benchmark results for CPU-accelerated models:

Note

Features in the benchmark result below: faceattr/age.v1.cpu, faceattr/beard.v0.cpu, faceattr/emotions.v1.cpu, faceattr/gender.v2.cpu, faceattr/glasses3.v0.cpu.

Number of instances	1	2	3	4	5
Model	RAM, Mb	RAM, Mb	RAM, Mb	RAM, Mb	RAM, Mb
face/ifruit_320.cpu	4300	5020	5698	7606	8279
face/ifruit_320.cpu + features	6254	7019	7725	8597	9220
face/ifruit_320.cpu + features + edie_rc2.cpu.fnk	6513	7454	9165	9685	10777
features	3034	3799	4460	5091	5774
features + medmask3.v2.cpu	3191	3942	4699	5313	6003
face/ifruit_320.cpu + medmask3.v2.cpu	4447	5108	5809	7642	8262
face/ifruit_320.cpu + edie_rc2.cpu.fnk	4583	5557	7746	8721	9592

Benchmark results for GPU-accelerated models:

Note

Features in the benchmark result below: faceattr/age.v1.gpu.fnk, faceattr/beard.v0.gpu.fnk, faceattr/emotions.v1.gpu.fnk, faceattr/gender.v2.gpu.fnk, faceattr/glasses3.v0.gpu.fnk.

Model	GPU RAM, Mb
face/ifruit_320.gpu.fnk	3169
face/ifruit_320.gpu.fnk + edie_rc2.gpu.fnk	4243
face/ifruit_320.gpu.fnk + features	3901
face/ifruit_320.gpu.fnk + features + edie_rc2.gpu.fnk	5102
face/ifruit_320.gpu.fnk + faceattr/medmask3.v2.gpu.fnk	3313
features	2345
features + faceattr/medmask3.v2.gpu.fnk	2617