Future of Facial Recognition

Author: John Honovich, Published on May 04, 2013

The Boston Bombings reaffirmed how far facial recognition needs to develop to be an effective surveillance tool. Despite the PR spin jobs, facial recognition failed to make any difference. Proponents, of course, are quick to talk about how fast improvements are coming and how all this will change in the future (ironically, just like they said 5 years ago). However, one Carnegie Mellon professor had insightful comments about the issues and needs for facial recognition. In this note, we break them down and analyze their likelihood of development.

People Look A Like

**** ****** **** * **** ** ************** ** ***** *** life ** **** ********* ******** **** *** ******. *******, * facial *********** ****** ** '******* **' **** *** **** ** people. ***** ** * ***********. **** ****** **** ******** ** visitors * ***, **** ** ********* * *****. **** **** many ******, ***** * ****** *** ***** ** **** *******. *** a **** ****** *******, *** ******* ** ********* ** ****** *** **** ***** *** *** not *******.

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

"*** ***** **** *** ******* ** ***** *********. **** *** start ******* **** ********* **** ******* ******** ** *****, *** start ** ******* **** **** ****** **** ******* ** **** other."

**** ******* ** *********** ******* **** ********* ****** ** ****** with ******** ********* *****, ********* ******** **********, **** ***** ** a ****** *****, ***. ******* ** ****, **** **** *** have **** ********* ** *****, *** *** ***** **** ********. Having **** ****** **** **** *** ** **** *** ******** the **** *** ******** ********** ********.

The ********?

*** ********* *** ** *********** ******* ** ***** **** *******, noting **** "***** ****** ** **** ** ************** ******* ***** is **** ** ****** ***’* ****** ** ****** *********** **** we ********* ***** ******. ** ** * ******** *********** ** the ******** ** *** ******, ***** ** *** **** *** dressed, *** **** *** *******, *** ** ***** **** *** geographically *** **********."

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

"****** *********** **** ***** ** *** ****** ******** **** ******** such ** *** **** *** ***, **** ** **** ******** way ** ********, ***** **** *** **** ****** ** ** in *** **** ** *****. **** * ****** *******, *** example, * *** *** *** ** ******* ** ***** *** connected ***** ***** ***, *** * *** ***, ******* ***** are ** **** ****** ** ***, *** **** *** ***, whether *** **** ****** ****** ** ***, **** ******** *****."

***** *** ***** ***** *** ******* **** ** ********* ******* but ***** *** **** ** ****** ******** **** ******* ****.

**********, ******, ***** **** ***** ****, ************ **** **** ***** be *********. ********* *******'* ****** **** ******* ** *** ******** without ****** ***********. **** ********* *** ******** ** ** ******* up ** **** ****** ** ******** **** ******* ***** *** hair ***** ** *** **** ****** **** ******. ***** *** be ****** ******* ** ********** **********, ***.

*********

*** **** ******** *********** *** ****** *********** **** ****** ** high ********** *******, ***** ******* ******** ****** ****** ** ******* quality ***** ****** *** ******* ***** ** **** ******. ****, at *****, ***** ** ******** ** ***** *****. *************, ** silver ****** ****** *** ********, **** *** ********** ****** ** correcting **** ** *** ************ ** ****** ******* **** **** other. 

Comments (5)

I don't see facial recognition being very succesful until the equivalent of a fingerprint is identified and incorporated into FR systems. Present day values such as the spacing of various parts of the face are insufficent for this purpose as they require both the subject and the comparison database images to be taken at approximately the same angle, elevation and distance to allow reasonably accurate comparisons.

For the technology to function accurately, some other means than standard 2D images will either have to supplant or enhance the means of identification. This could be heat patterns, under-skin structure (arteries, veins and cappliaries) iris prints, pore locations or some other more unique and less concealable criteria.

I worked on NEC's NeoFace Facial Recognition Software for 5 months, and it is worth to give a chance that software.

Can you expand on your use of it? Anything specific that they did that stood out? How was it used?

Does facial recognition software exploit the 3D information inherent in video?

As a person passes through a surveillance field of view, both horizontal and vertical aspects change significantly, providing the raw information necessary to construct a 3D model.

There is a well developed body of literature that addresses mechanisms to expand multi aspect observations into a 3D model of elements of interest within the scene.

The literature also addresses mechanisms to integrate deformable maps into a consistent 3D model (facial expressions may change, persons may be speaking, etc. which increases the challenge of extrapolating from multiple frames into a ocnsistent 3D model).

A more complex 3D model could increase points of difference across similar populations as well as reduce aspect dependence.

Another complication the article did not address is the fact that people's faces are asymetrical, so for example a left view may differ substantially from a right view. Of course, one cannot accurately reconstruct elements not present within the observed scene, again limiting the dimensionality of the match.

This article is a little old, but Facial Recognition stills as a interessant technology to many applications.

What about 2018... does anyone have a up-to-date opinion? What can we expect about the present point of evolution?

I have heard about GPU processing, 3D face recognition, rotation compensation, etc... How this technologies can improve the accurance and what is the impact in infraestructure?

Recently I saw a article from Ayonix launching an Axis's cameras embedded 3D facial recognition. They claim that their solution has a 99.X hit-rate using a single server from 10.000 cameras... Did anyone test this?


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 Facial Recognition

GDPR For Video Surveillance Guide on Apr 12, 2018
The European Union’s General Data Protection Regulation (GDPR) comes into force on May 25, but there is much confusion and no clear guidelines on...
Destructive Video Analytics Hype Returns on Mar 27, 2018
It is not just Hikvision's false advertising campaign. With marketing money being pumped into deep learning, we are returning to the bad old...
Why 3VR Failed on Feb 16, 2018
3VR destroyed transformed ~$65 million in VC funding into a $6.9 million exit. The reason they failed is simple. They bet on analytics. They...
Chinese Police Wearing Facial Recognition Are Here on Feb 06, 2018
This is a very interesting and highly atypical usage of facial recognition that the Chinese government touted this week: It is a PRC police...
Chinese Government Attacks Western Reports on Jan 03, 2018
The Chinese government is angry at the BBC and WSJ's reporting on Chinese video surveillance (see BBC Features Dahua and WSJ Investigates China's...
2018 Top Sales Interest For Integrators on Dec 20, 2017
130+ integrators answered: This is a key question, especially for manufacturers, as it impacts what areas to focus more on and what...
BBC Features Dahua on Dec 13, 2017
Hikvision is not the only mega-Chinese video surveillance manufacturer getting global attention. Last month, the WSJ investigated Hikvision and now...
Lighthouse Deep Learning Camera Tested on Dec 07, 2017
A Silicon Valley startup, Lighthouse, with a Stanford PhD CTO, has released a deep learning AI camera with 3D sensors for just $300. The company...
BBC Featured Deep Learning Face Recognition Analyzed (Digital Barriers) on Aug 30, 2017
The UK's largest broadcaster, BBC, recently featured Digital Barriers (also based in the UK) on a segment highlighting face recognition technology....
Nest Cam IQ Tested on Jul 10, 2017
Nest has released their latest entry in their camera line, the Nest Cam IQ, touting 4K "Supersight", facial recognition, "HD audio", invisible IR,...

Most Recent Industry Reports

May 2018 Camera Course on Apr 20, 2018
Save $50 on early registration until this Thursday, the 26th. Register now (save $50) for the Spring 2018 Camera Course This is the only...
Global Real-Time Video Surveillance - EarthNow on Apr 20, 2018
A new company, EarthNow, with backing from Bill Gates, Airbus and more, is claiming that: Users will be able to see places on Earth with a delay...
Dedicated Vs Converged Access Control Networks (Statistics) on Apr 20, 2018
Running one's access control system on a converged network, with one's computers and phones, can save money. On the other hand, hand, doing so can...
April 2018 IP Networking Course on Apr 19, 2018
This is the last chance to register for our IP Networking course. Register now. NEW - 2 sessions per class, 'day' and 'night' to give you double...
Rare Video Surveillance Fundraising - Verkada $15 Million on Apr 19, 2018
Fundraising in video surveillance (and the broader physical security market) has been poor recently. Highlights are few and far in between...
'Best In Show' Fails on Apr 19, 2018
ISC West's "Best In Show" has failed. For more than a decade, it has become increasingly irrelevant as the selections exhibit a cartoon level...
Security Camera Cleaning Frequency Statistics on Apr 18, 2018
150+ integrators told IPVM how often they clean cameras on customer's sites and why.  Inside we examine their answers and break down feedback...
Worst Access Control 2018 on Apr 18, 2018
Three access control providers stood out as providing the most problems for integrators. In this report, we analyze the answers to: "In the...
Axis VMD4 Analytics Tested on Apr 17, 2018
Axis is now on its 4th generation of video motion detection (VMD), which Axis calls "a free video analytics application." In this generation, Axis...
Arecont CEO And President Resign on Apr 17, 2018
This is good news for Arecont. Arecont's problems have been well known for years (e.g., most recently Worst Camera Manufacturers 2018 and starting...

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