Customize our computer vision solutions into your infrastructure or product.

CPU Benchmarks

Speed x86

Speed ARM (RPi 4)

Ram Req.

Face recognition lite model

5.5 ms

35 ms

55 MB

Face recognition full model

75 ms

1200 ms

2 Gb

Face recognition 1 : N search (N = 1,000,000)

Face and Landmark detection

10 ms

12 ms

66 ms

70 ms

750 bytes / template

25 MB

Object detection

10 ms

85 ms

54 MB

Liveness

54 ms

340 ms

56 MB

*All benchmarks performed using 1280x720 pixel images containing 1 face or object with CPU only with smallest face height set to 40 px. x86 benchmarks were performed on dual Intel(R) Xeon(R) CPU E5-2630 v4 @ 2.20GHz, 128GB Ram Ubuntu 18.04.4. RPi 4 benchmarks performed using 1.5GHz 64-bit quad-core ARM Cortex-A72, 4GB RAM, Gentoo Base System release 2.6 (AArch64). Ram usage refers to maximum resident memory. 1:N identification speed and memory benchmarks recorded using full model with frVectorCompression flag enabled. Enrollment template size represents conservative average case, but it can be variable due to variable length in identity string.

GPU Benchmarks

Average Time per Input

VRAM Usage

Face and landmark detection

4.6 ms

1.9 Gb

Face recognition TFV5, batch size = 1

8.7 ms

2.73 Gb

Face recognition TFV5, batch size = 4

Face recognition TFV5, batch size = 16

4.1 ms

2.62 ms

3.22 Gb

4.91 Gb