Amazon Rekognition Facial Detection Tested

Author: IPVM Team, Published on Dec 03, 2018

Amazon's Rekognition service has announced improvements, that if accurate, would be a significant advance over conventional facial detection.

amazon recognition

In particular, Amazon claims:

Amazon Rekognition can now detect 40 percent more faces – that would have been previously missed – in images that have some of the most challenging conditions described earlier... These aspects might include pose variations caused by head movement and/or camera movements, occlusion due to foreground or background objects (such as faces covered by hats, hair, or hands of another person in the foreground), illumination variations (such as low contrast and shadows), bright lighting that leads to washed out faces, low quality and resolution that leads to noisy and blurry faces, and distortion from cameras and lenses themselves. [emphasis added]

IPVM developed a dataset of 100+ 3-second video clips:

facial detection dataset 2

These clips are done across a number of common real-world scenarios:

  • Day light (200 lux)
  • Low light (1 lux)
  • Side tilt (from 0° to 90°)
  • Down tilt (from 0° to 30°)

We compiled various combinations (such as 200 lux, 15° side tilt, 22° down tilt or 1 lux, 75° side tilt, 30° down tilt, etc.).

We then ran tests of these 100+ clips across facial detection approaches:

  • HAAR
  • HOG
  • Openvino (specifically the face-detection-retail-0004 model on Intel i7 and Myriad X chips expanded upon from our NCS2 test)
  • Amazon Rekognition

Inside, we explain full results and key variances in performance.

******'* *********** ******* *** ********* ************, **** ** ********, ***** be * *********** ******* **** ************ ****** *********.

amazon recognition

** **********,****** ******:

****** *********** *** *** ****** ** ******* **** ***** – that ***** **** **** ********** ****** – ** ****** **** have **** ** *** **** *********** ********** ********* *******... ***** aspects ***** *******pose ********** caused by head movement and/or camera movements, ************ ** ********** ** ********** ******* (**** ** ***** ******* by ****, ****, ** ***** ** ******* ****** ** *** foreground),illumination ********** (such as low contrast and shadows), bright lighting that leads to washed out faces, low ******* and resolution that leads to noisy *** ****** *****, and distortion from cameras and lenses themselves. [emphasis added]

**** ********* * ******* ** ***+ *-****** ***** *****:

facial detection dataset 2

***** ***** *** **** ****** * ****** ** ****** ****-***** scenarios:

  • *** ***** (*** ***)
  • *** ***** (* ***)
  • **** **** (**** *° ** **°)
  • **** **** (**** *° ** **°)

** ******** ******* ************ (**** ** *** ***, **° **** tilt, **° **** **** ** * ***, **° **** ****, 30° **** ****, ***.).

** **** *** ***** ** ***** ***+ ***** ****** ****** detection **********:

  • ****
  • ***
  • ******** (************ *** ****-*********-******-**** ***** ** ***** ** *** ****** X ***** ******** ******** *** **** ****)
  • ****** ***********

******, ** ******* **** ******* *** *** ********* ** ***********.

[***************]

Key ********

****** *********** ****** ********* ******** *** *** ****** **** *** others ****** ********* ****** ** **** ** ********:

facial detection overall accuracy

*** ****** *** ********* ***** '***' *** **** ** *** examples ** **** **** ***** ****, ********* *** ********* *** bad ******.

*** *******, **** ** ******* *** ****** ** *** ********** used, **** *** ****** ***********, ***** *********** ** *********** ******* are *******:

rekognition accuracy segmented

**** ******* *** *** ****** ****** ******** ***, **** '****' light *********** ** ****** **% ***** *** ***** *********** ** under **%.

***** *********** *** ********* ** ****** ********* ********, ** ***, far *** ****, *** *******, ** *** ****** ** ********* versus ******** *** **** ******* ** ** **:

facial detection fps

*********** ** **** ****** **** *** ***** ******** ** *** Rekognition ***** ******* ******** *** ***** ** ** **** ***** to *** ***** *** ***** ******** *** ** **** ******* / ** *** **** / ***** *** ****** **. ** examine ***** *********** ******* ** *** ***** ******* ** **** report.

********

*****, ** ******* ** ***% ******** *** ****** ***********, ***** is ****** ****** *********** *** **** ** **° **** *** 45° **** **** **** *** ******:

[video-to-gif output image]

****, ** ******* ** **% ******** (******* ~* *** ** 10 ****** ********* ******* * ****) *** ****** ***********, ****** a ***** ********* ** ***** **** *** **** **** **** the *******'* **** ****:

*** ***** ******, ** *****, **** **** **** *********** **** for ***********. *** *******, **** * *** ** ***** * frames *** * **** ********** ******** ****** **** * ***** angle *** *** ****** / ********:

*******, *** ** *** *** **** ************ *** *********** *** in **** **** ********, ****** *** *****, *** ******* ****** straight ******* *** ****** *** ***** **** *** *** ******** at ***:

Accuracy ***********

*** ******** ****** **** ** ************** *** * **************. *********** face ********* ********** **** * ****** ******"**** ******* *********" (***), ***** ******** ***********-**-******* ** ****** **** ********* ******** ************** ** ****** **** ********-*** ******* (********* ****** "************ **** Union" ** ***) *** ******* **** *** ********* ******** * face. ** ****'* ***** *******, **** **** ****** *** *** face (***** ***** *** * ****, *** ***** ** **** one). ***** ****, ** *** *** * ****** ********* *** accuracy: ** *** **** ****** *** * **** ******** (******** / *****-***** = ********). **** ** *** *******: ** **********, it **** *** ******* *** ***** ********* (* **** ** detected ***********, **** ** *** **** ** ** *** *******'* shirt). **** ******* ****** - ** ** ***** **** *** heuristic. ** ****** ** **** ******* **** ********** (***, *****, etc) ******* ****** ** ***** ******** ***** *** ***** ***** of *** *******.

FPS ***********

*** ******* **** ****** **** *** *********** ***:

  1. ****** *** ***** ** ****** **
  2. ***** *** *****
  3. ***** *********** (******** ** ** *** ** ****, *********** *****->***********)
  4. **** *** *****, ********* ***

* **** ****** ******** *** ************ ***** ** ** ****** videos ******** **********->***********. **** ***** *** *** ********** ******** ** * *****-**-***** basis, ***** *** ***** ***** *** ********** ******** **** *** FPS *********** ********. ** ****, *** *** ***********, ******* **** low, ***** ** **** *****.

***** ** **** ************, **** ** * ***** ***** ******* Rekognition, ** ****** ** *** ** * *******. *** *******, one *** *** ***-***** ******** *** ****** ******* ***********. *** other ****** ** **** ******* *** ** ****** * ** Intel ** ******* (****-***) *******, ** **** *********** *****-*** ** situation-dependent. ************, *********** ** * ******* *******, ***** *********** * model (**** ********'* ****-******-****) ***** **** ************* *** *****. ** while ****-**** *********** **** *********** ** ******** (****: ** *** still ************* ** **** ** ********), *** ***-****-**** ************, *********** could ** ***** ******.

*******

*** * ****-**** ***** ************ ***********,*********** ************ ** ***** **** *** *** ***** ************ ***** **** infrequent ***** ********** ** ******, ** ***** ** *** **** affordable **** **-**** **********. *** *******, *********** ******* *** **-**** (N. ********) ** $*.** *** * *** ** **** ****** video ********. **** **** **** **% ******** ** ******* (*.*., 4,380 ******* ** * *****), *** ******* **** ***** ** ~$50 (*,*** * **% * $*.**) **** *** *** **** analytics.

** *** ***** ****,****** *********** ****** ******* ***************** ****** ***********, ****** ******** (***, ******, *******), ****** *** activities *********, ***. **** **** ********* ******** ****** ** *** have ** ***** ** ****** ** ******** ** ******* **** their ***********.

********

**** ** *** *** ** ****'* *** ****** ** **** Learning *****, ******** ********** ********* ******,***** ****** ******* ***** * / ******** ** ******* *** ****. ** *** ******** * ******* ** *** tests ** *** ******** ******. *** ********, ******** ** *********, please *** ** *** ********.

Comments (7)

John Honovich

IPVM | 12/03/18 01:52pm

*** ***** ** *** ********** ** *** ************,***** *** *** *** ***** **** ****:

deep learning webinar2

***** **** ** ********** ************ ** **** ** *** ******** effort ** ******* ************* ******** *** *********, ** ******** ** our ******* *****.

Undisclosed Manufacturer #1

12/03/18 02:50pm

*** *** ******* **** ******* *** ********** **** ** *********** on? **** ******* ***** ****** ** ** *********** (** **** have *** ******) ** * ***

John Honovich

IPVM | In reply to Undisclosed Manufacturer #1 | 12/03/18 09:02pm

** *** *** ** *** ** ******** *** ** **** try **** ** ****** *****.

Avatar

Matthew Del Salto

BLACKBOX USA | 12/04/18 12:12am

**** **** ***** **** ********* ** **** **** ** **** haven't *******.

Undisclosed #2

12/03/18 07:45pm

** ***** *** *** *** *** **** *** ** **** raw ***** **** ** *** ***** *** **** ***** ** download *** *** **** ***** *** ********* *******?

Undisclosed Manufacturer #3

In reply to Undisclosed #2 | 12/03/18 08:02pm

* ***** **** ***** ** * ****** ***** *** ** compare ********.

John Honovich

IPVM | In reply to Undisclosed #2 | 12/03/18 09:03pm

** ***** ******* *** ******* *** **** *** ******* ***.

Login to read this IPVM report.
Why do I need to log in?
IPVM conducts unique testing and research funded by member's payments enabling us to offer the most independent, accurate and in-depth information.

Related Reports on Low Light

Startup Spectral Edge RGB+IR Claims "Unique Low Light Performance Tech" on Apr 03, 2019
This UK startup, born out of university research, is boldly proclaiming that they deliver "Unique low light performance tech for video surveillance...
Axis 5MP P Vs Q Camera Shootout (P3227-LVE vs Q3517-LVE) on Apr 02, 2019
While Axis' Q3517-LVE tested well, it is one of Axis' most expensive dome cameras, over $1,000 USD. Because of this, members asked us how it...
Covert IP Camera Shootout - Axis, Hanwha, Hikvision, March, Vivotek on Mar 20, 2019
Covert cameras were one of the last holdout areas for analog cameras. However, in the past few years, IP / HD covert cameras have become...
Axis IR Multi Imager Camera Tested (P3717-PLE) on Feb 18, 2019
Axis has released their first IR multi imager, the P3717-PLE, a repositionable model listing 360° IR illumination and flexible positioning,...
Hanwha Smaller Multi Imager Tested (PNM-9000VQ) on Feb 14, 2019
Hanwha's first repositionable multi imager PNM-9081VQ tested well, but was huge, over 12" wide and weighing in at over 10 pounds. Now, they have...
Bandwidth vs Low Light Shootout - Avigilon, Axis, Bosch, Dahua, Geovision, Hanwha, Hikvision, Uniview, Vivotek on Feb 08, 2019
Nighttime bandwidth spikes are a major concern in video surveillance, but do all manufacturers' cameras perform the same? Are some more consistent...
Sony Favorability Results 2019 on Feb 06, 2019
Sony Favorability amongst integrators improved moderately compared to their 2017 favorability results, with a modest net positive...
Camera Course Winter 2019 - Last Chance on Jan 24, 2019
This is the last chance to register for the Winter 2019 Camera Course. This is the only independent surveillance camera course, based on in-depth...
Testing Bandwidth Vs. Low Light on Jan 16, 2019
Nighttime bandwidth spikes are a major concern in video surveillance. Many calculate bandwidth as a single 24/7 number, but bit rates vary...
WDR Tutorial on Jan 11, 2019
Understanding wide dynamic range (WDR) is critical to capturing high quality images in demanding conditions. However, with no real standards, any...

Most Recent Industry Reports

H.265 Usage Statistics on Apr 19, 2019
H.265 has been available in IP cameras for more than 5 years and, in the past few years, the number of manufacturers supporting this codec has...
ACRE Acquires RS2, Explains Acquisition Strategy on Apr 19, 2019
ACRE continues to buy, now acquiring RS2, just 5 months after buying Open Options. One is a small access control manufacturer from Texas, the...
Access Control Course Spring 2019 - Last Chance on Apr 19, 2019
Register for the Spring Access Control Course. IPVM offers the most comprehensive access control course in the industry. Unlike manufacturer...
Riser vs Plenum Cabling Explained on Apr 18, 2019
You could be spending twice as much for cable as you need. The difference between 'plenum' rated cable and 'riser' rated cable is subtle, but the...
Verint Victimized By Ransomware on Apr 18, 2019
Verint, which is best known in the physical security industry for video surveillance but has built a sizeable cybersecurity business as well, was...
Milestone Drops IFSEC on Apr 18, 2019
Milestone has dropped out of Europe's largest annual security trade show (IFSEC 2019), telling IPVM that they "have found that IFSEC in EMEA no...
The Fastest Growing Video Surveillance Sales Organization Ever - Verkada on Apr 17, 2019
Verkada has the fastest growing video surveillance sales organization ever. In less than 2 years, they already have more salespeople in the US...
Door Operators Access Control Tutorial on Apr 17, 2019
Doors equipped with door operators, specialty devices that automate opening and closing, tend to be quite complex. The mechanisms needed to...
Securadyne CEO: IPVM 'Entertaining For An Ignorant Few' on Apr 16, 2019
Securadyne's CEO Carey Boethel is unhappy with IPVM's report - Failed Integrator Rollup, Securadyne Sells to Guard Giant Allied. Indeed, he...
Dahua Repositionable IR Multi-Imager Camera Tested on Apr 16, 2019
Dahua has released their first repositionable multi-imager camera, the Multi-Flex 4x2MP, claiming integrated IR, true WDR, and flexible...

The world's leading video surveillance information source, IPVM provides the best reporting, testing and training for 10,000+ members globally. Dedicated to independent and objective information, we uniquely refuse any and all advertisements, sponsorship and consulting from manufacturers.

About | FAQ | Contact