Simplicam Facial Recognition Tested

Avatar
Derek Ward
•Published Jan 23, 2015 05:00 AM

****** ***********, ********* *** $***?

****'* *** ***** **** * *******, Simplicam, [**** ** ****** *********] *** *** not **** ****** ******* ***** *** user ********* *** *** ***** ** facial ********* *** ***********.

** **** ******, ** ***** *** test ******** ** *********, ******* ** against *******:

** **********, ** ******* *********'* *** differentiator, ****** ********* *** ***********.

  • *** ***** **** *** ****** **** registered *****?
  • **** ** ******* *** **** *** the ***** *** ***?
  • *** **** ***** ****** ***********?

    • **** *** *** *** ******** **** our *****:

    ****

    • ** *** **** ******* *** ***'* want****** ** ***** **** **** ************ video, ********* *** ************ ****** *** the *** ***** *** **** ********* elements. ** ***** ****** **** ** OEM.
    • ********* *** ***** * ********* / event **** **** (**** ******* **** not ****) ***** ***** ** ****** to ****** ******** *****. *******, *** thumbnails *** ******** ** ** *** be **** ** *** ******* ******.
    • ****** ********* (*.*., *********** * **** but *** * ******** ******) ****** ~90% ** *** **** **** ****** were ****** ******* *** ******.

    ****

    • ********* ********** ******** ******** **** ***** **% of *** ****, ******* ***** ********* within ************ *************** (~*.*' ******** ******, ~*-*' to ******).
    • ***** *** ******'* **** ***** (***°) and ***** *********** ***** (*-*'), ** was **** ** *** * **** face **** ******* *** ****** *********** walking ** ** *** ******.
    • ******* *** ****** ** ~*' ****** ******* ********* in ****** *********** ********
    • ******** *********** ****** **** ** *********** *** faces ******* **** *** ******* *******, mistaking *** **** *** *******, ** strangers ** ******** *****
    • *** *** ********* **** ** ***********, forcing ***** ** ****** ****** ******* before **** *** ***********.

    ****

    ***********, *********'* ****** *********** ** ** beta, ****** ********* ** *** *********. We ****** **** ** ******** ** be ***** *** *** ********* ***** how **** ****** *********** *********** *** be ********, ********** ***** *** **** wide ***.

    *****

    *** ********* ***** *** $*** ****** [link ** ****** *********], *** ********* a ********* ****. ********* ***** ***** at $**/**** *** *** *** *********, $129 *** ** **** *********, *** $229 *** ** ****. ********* ***** are **% *** *** ******* ****** the ***** ***.

    ** $***, *** ********* ** $** less ********* **** ********** ***. ******* ****** * *** ($**) and ** *** ($***)********* *****.

    ***************

    ***** *********'* **** ****** *********** ********, ** should *** ** ****** **** ** home ** ********** ********. **** ********* recognized ** *****, *** **** *****, along **** ****** ******* **********, ******** false ********* *** ****** *********** *** likely.

    *** ***************-******** **** ***** *** *** aware ** ***** *********** *** ******* to **** **** ****, *** ********* may ** **********, *** *** "***" factor, ** **** ** ********* ** alerts ** **** *** ******** **** by ****** ***********.

    ** * ******* ***********, ********* ** the ******* ** **** ***** ** consumer *******, ** ~$** ***** ***** point. *******, ***** *********** ********* **** is ******* ** * ****** *** of *******, ******* ***** ** ****** pay **** **** ****** ($*** *** 11 ****) *** ******** *******, ** react ** ****** ******* ****** ***** is ***********.

    ******** ********

    *** ********* ** ***** ** **** to *** ******* ***, ****** ******** larger ** ***** ********. ** *** be ****** ****/***** ******* ** **** mounted **** *** ******** **** ****.

    Facial ***********

    *********'* ****** *********** ******** ***** ***** to ****** ** ** ** ******, a ~** ****** *******. **** ********, facial *********** ****** *** ***** ** the ********* ******** *** ***** ****, with ********** ***** ********* ** ******** video *** **********. ***** *** **** alerts ("**** *** *** ******** ** 1:38pm") *** **** ** ********* ** recogntion.

    ********* ********** *** ****** ** ******* about *-*' ****, *** ******** ** no **** **** *-*' ****.

    ********* ** ******

    ** ****** ********** *********** ******, ***** this *****:

    *********** ********

    ** **** ****** *****, *** ******* was ******** ********, ***** *-*' **** from *** ******, ***** ** ******* at ***** *.*' ****. **** *** looking *******, *** *** ***** **** lit (~*** ***), *** **** ** clear.

    *******, ** *** **** *****, **** a ****** ******** **** **** ******** recognition:

    ******* *** ****** ** ~*' **** decreased *********** ***** ** ****, ** about **-**%. 

    ***** ******

    ** ******** ** ****** ***********, *** camera ********* ******* *** **** *** another, ** ** ************ ****** ** a ****. **** *** ********** ****** when **** **** *******.

    ************, ****** **** ******, *** ****** regularly ******** ******** ******, ******* **** one ***** *******. *** ******* ** the *** **** ****** ** *** image ***** ******** ****.

Comments (7)
Avatar
Scot MacTaggart
•Jan 23, 2015

...now imagine how many failures you'd have in a residential setting, where family resemblance would become a factor.

(3)
JB
Jeremiah Boughton
•Jan 23, 2015
In reply to Scot MacTaggart

I would recommend not turning that setting on. That setting is just a selling point with geeks that don't know about security cameras and the true state of video analytics.

Avatar
Derek Ward
•Jan 23, 2015
Hanwha

We contacted Simplicam support asking what they recommend for enrolling subjects that wear glasses regularly, and they responded with "Enroll them with what they would normally wear on a day to day basis".

We told them of the high missed or incorrect facial recognition detections with subjects enrolled with glasses, and support replied with "Thanks, we were not aware of that. We will look into it more as we move it out of beta".

JB
Jeremiah Boughton
•Jan 23, 2015

Simplicam has a major feature that DropCam is completely missing. That main feature, when I tested for 2 weeks, was the Facial Detection (non-human versus human) and alerts. This helped lower the alerts from cats, dogs, cars passing by and from Sun movement. DropCam is a major pain in these areas. With a few Central Stations now offering Video alarms/verification, Simplicam would be more favorable with Central Stations. (Central Stations can get email alerts with clips into their systems now)

Still both DropCam and Simplicam run into problems with residential families with Upload bandwidth when they get more than 1 camera. The average basic High-Speed Upload is around .75 Mbps (+10 -10) in Wisconsin (Download is usually pretty high 13 to 40).

Both of these cameras use 400-500K on average and average 1Mbps with a couple motions at 1 time. More than 1 camera is a great way to make the average residential upload come to a crawl. Mom and Pop Businesses roughly average 1.5 to 3Mbps for upload times and if they go up a tier they usually get 10-20 for upload (Again Download speeds are not a problem)

Residential customers would possibly have to get a little network setup savy if they go with a few cameras due to using cheap residential hardware. They should probably have their own switch and Access Points if they go with 4+ cameras just to prevent possible internal bandwidth jamming (family watches stored movies from NAS and streaming music)

On a side note - SimpliCam has did a great job at Marketing to the Residential Market. Every other day SimpliCam is in my Facebook newsfeed. DropCam was more about viral marketing through various Tech blogs in my opinion.

On a side note - I have two DropCam Pro's and 1 Simplicam and a few other cameras that I have tested. Currently they are all laying around collecting dust, as I really do not have a strong enough need for them. I'm looking for something that stores locally and allows me to view remotely with an app and get alerts. I'm currently testing some cheap Chinese cameras, which look promising so far, with a decent UI and local storage. The User Interface (VMS) usability is important to me, hence why I'm not a fan of the Foscam.

(4)
(1)
TC
Trisha (Chris' wife) Dearing
•Jan 24, 2015
IPVMU Certified
In reply to Jeremiah Boughton

I'm looking for something that stores locally and allows me to view remotely with an app and get alerts.

Have you looked at the iCam360, in this thread?

(1)
JB
Jeremiah Boughton
•Jan 26, 2015
In reply to Trisha (Chris' wife) Dearing

I have not tested this cam yet. It seems to lack video management. It records and stores, however it does not have a management system.

(1)
KA
Konstantin Avramenko
•Jan 24, 2015

As a long term advocate of usage FR within CCTV I can just repeat my point: use the right tool for the right purpose. I did not test this product but it is obviously does not fit with the used test scenario. From the first view it can be substitute for cases where people now use web cameras: cooperative, close range identification or even better verification.

FR cannot provide with the same level of accuracy as for example iris or vein pattern recognition. But so far it is the only modality that can provide with the result on a distance. If your customer has watchlist bigger than 100 suspects it is just impossible to expect to recognize them manually (security officers with this ability are already in a CIA). For example, a project where the watch list consists of 2K suspects, throughput is around 100 - 300 people per hour, the user gets up to 100 alerts per day and on average only 15% of them are true match. The hardware and software are top notch, the watchlist material is pretty bad but it is usual thing. At the same time the customer is satisfied. Why? Because they get something instead of nothing. They can manage the load of their staff and get predictable results in a long term run.

So, use the right tool for the right purpose and get benefits from the biometrics.

(1)
(1)
(1)