Gun Detecting Smartcam Startup

By Brian Rhodes, Published Jul 27, 2015, 12:00am EDT

Wild claims are nothing new for security products, but one startup is claiming its home camera can detect guns.

GunDetect claims it can stop gun crime and accidents in your house cold by sensing guns it can see. The new company is so confident in its claims, potential customers can try it out for free [link no longer available]. So we did.

 

In this note, we look at the product and see if they can deliver.

Concept ********

*** ***** ***** **** the ************ **** ******** the *******

Key ******

***** ******* *** *** to ** ********* ********** on * ********** ********, certain ******* *** ****:

  • ****** *** *********:*** ****** ****** *** precence ** **** ******* object ***********. *** ******* claims *** ******* ********** a ***** ****** ** firearms,"******* **** ********** ** machine ****. ********* ** optimized ** ****** **** in *******’* *****, ******* of ********** ********** ** a **** ****** * gun ** * ***** or *** ** *********."
  • *** ********:  *** ******* **** be ******** ** *** variants, ** ****** '*****' or '*******' *******. *** $399 ***** ******* ***** video ** ********* ******* offsite **** ***** ***** for **** *** ***** alerts ** **** ***********. The **** ********* ******* option **** ******* ****** in *** ******** ********* locally *** **** *** same ******. *******, *** $599 ******** ***** ****** compared ** *** ***** version, ********** ******** ** firearms ******.
  • ****** ******: **** * ******* is ********, ****** ****** uses *** **** ****** to ****** *****. * cloud-based *** ****** ** alerted ** *** ***** camera *** ********, *** the ****** **** ****** a **** ******* ** remote *****.  ** *** is ******.

Product ********

** **** *********** ******* or ***** ** *** camera **** **** ********, but *** ***** ******* will ** *** ******:

*** ******* ******* ***** the ****** ********* ** a ******** ********* ****, depending ** *******, ** the ***** ********* ********* or *** ******* **** interface **** ***** ********* servers.

***** **** ******* *** appliance ** *********'* ***** Analytics ** *** ******* through * *** ****'* home ****.

*******

*** *********** *******, *** **** ********* 'Cloud' ******* ***** $***, and *** ******-********* '*******' version ***** $***.  **** camera ** ************* ******** and *******, *** * user **** ******** ***** will **** ** ******* them **********, *** *** additional ******* ** ***** to *** ********* **** appliance.

Visual ********* ****

****** ***** ******** ********* that *** ***********, ********* uses * ******* ****** system ** *** ********. The ******* ****** **** is *** ** ******* for *** *********** ******, as ******* ***** ****-********* scanning ** ********** *** significantly **** *********, **** even ***** ******* ****** systems ******* $*,*** *** **** *** camera. 

*******, ********* *** ****** 'see' *** *** ***** that ** **** ** unable ** ****** ********* or ********* ******** ********, even ** *** ****** is ******* *** *** the ********* ** **** in *** ******'* ***** of ****.

Free **** **** 

** * **** ****, GunDetect ******* *** ** Try ****** *** *** [link ** ****** *********] and ********** ***** ** upload ***** *** ******** to *** ** ***** analytic *** **** **** in ******.

** ********* * ********* photos ** **** *** engine, ** ******* ****** that ******** **** ********.  Of *** * ** submitted, ********* **** ********** ID'd *** *** ***** *** failed ** ***** ******* scenes:

*) ** ***** **** robber ******** * *******:

 *) * ******* *********** a **** ******* ****** the ******:

** **** *****, ******* a ********** ***** *** distinct ******* ***** ** the ***** ** ****, the ******** *** *** detect * ***.

** ******** ******* ***** with * ******, *** distinctly ****, ******** ****** several ***** *** ****. *** analytic ****** **** ****:


***** **** ** ****** between **** *** **** guns ** * ******* strength ** *** ********, with *** ******* ****** "**’** *** ***** ******* training ** ** ****** water **** *** *** gestures **** **** *******. For ********* *** ****, such ** *** ******* pistol ** **** **** the ***** ****** ** our ******** *****, ** is ******** ****** ** err ** *** **** of ******." *** ***** ******* with **** ********* ** instead ** '****** ** the **** ** ******', it *** *** ***** at ***.

*******, *** ******** *** not ****** ****.  *** example, *** *** *** below *** *** **** the ******:

*** *** ****** *** properly ***** ***** ** the ******* ** **** scene:

*******, *** ********** **** detection ******** ** **** and **** **** *** '**** **%' ******** *******.

Our ***************

******: **** **** **** GunDetect.  **** ** *** system ** **** ** identify * ******* ** your ****, ***** ** a **** ****** *** FoV **** ****** ** too ***** ** *** gun **** ** ********* obscured ** **** ****** for *** ***** ** work.

** ********* *** ****** is *** ********** *******,****** * ****** *** safe ** ********** ******* ****** ***** from ******, *** * large ***** ** **** hard *** ******* **** less **** *** ****** price *** *********.

Comments (13)

Avatar

John Bazyk

Command Corporation
IPVMU Certified | 07/27/15 02:13pm

I agree with your recommendation. Although $500 for a gun safe is a little on the low side. We install alarm contacts on safes all the time and have them send push notifications to the users phone when the safe detects vibration or its opened. User can then check their cameras using video verification during the alarm. Much better solution.

John Honovich

IPVM | In reply to John Bazyk | 07/27/15 02:15pm

"$500 for a gun safe is a little on the low side."

Big, big volume discounts in Oklahoma...

Avatar

Brian Rhodes

IPVMU Certified | In reply to John Bazyk | 07/27/15 03:20pm

Good point, worth clarifying. 'Gun safe' is a fairly broad category.

I think you're talking about the vault type, with the formal spinning combo lock, fire rated heavy steel construction, and 15 or 20 gun capacity. Indeed, I'll agree $500 does not fairly describe those. Maybe not even $5000 for some of the 'deluxe' models that have more options than a new Cadillac. Really nice! (Although I've never seen a 'living room' install):

I am referreing to a handgun safe, which is generally small enough to cram under a bed. A small box like that is easily stolen and even pried open with tools, but it does keep small children from handling the contents.

Avatar

John Bazyk

Command Corporation
IPVMU Certified | In reply to Brian Rhodes | 07/27/15 04:50pm

I figured you were referring to the small lockers. I've got the same Liberty. No ones picking it up and moving it. Even with those small safes you can throw a wireless transmitter on, that will give you better detection than one of those gun cams.

Avatar

Jonathan Lawry

Trecerdo, LLC
In reply to Brian Rhodes | 07/28/15 01:29pm

I actually have seen a living room install. A paramour of mine from my younger days had one in her living room, filled with what must have been over a hundred grand worth of antique firmarms and handguns. She was a champion trapshooter, in spite of being of very petite stature.

Every time I went to her apartment, that Tommy Shaw song, "...I thought I could handle a girl with guns..." would go through my head.

The relationship probably lasted a few months longer than it should have. Good times.

Undisclosed #1

IPVMU Certified | In reply to Jonathan Lawry | 07/30/15 08:48pm

Surely you got to see her BedBunker? ;)

Stephen Schulz

IPVMU Certified | 07/28/15 02:45pm

If I roll my eyes any more I will be blind.

C'mon, is this a real product?

John Honovich

IPVM | In reply to Stephen Schulz | 07/28/15 02:48pm

Nothing on Kickstarter is a real product :)

Avatar

Billy Guthrie

07/29/15 12:41am

I think they would be better licensing the software out; I see more benefits using the analytics in many different other ways rather than in a single product. About the testing, sure, the analytics could use some work; the 40+ images that I used, 36 of them accurately detected the gun, even at a distance of 15 feet away.

Avatar

David Warren

Security Network Advisors LLC
07/30/15 12:55pm

Not sure what's worse, this idea or Tom Smykowski's Jump-to-Conclusions mat.

Jason Spielfogel

08/04/15 04:19pm

Not a bad idea, but you're going to need a false-negative rate at least below 10% for this to be a viable commercial analytic.

No storeowner will buy this if 60% of the time it fails to detect a gun

Chris Dearing

IPVMU Certified | 08/04/15 04:54pm

You are being more than fair in granting that the toy gun did not 'fool the system'. Based on the other results, I doubt it even considered the toy gun as a candidate weapon.

But I'm wondering why the thing is using still photos to detect firearms and not several frames of video? Object recognition is greatly aided by even just a few extra frames.

Also, maybe there is a glimmer of hope in the fact that it asks you if it guessed right or not, does that mean that it is learning and will do better in the future?

If it learns from its mistakes, and since criminals often return to the scene of the crime, you may have better luck with repeat offenders. ;)

Avatar

Brian Rhodes

IPVMU Certified | In reply to Chris Dearing | 08/04/15 05:59pm

Also, maybe there is a glimmer of hope in the fact that it asks you if it guessed right or not, does that mean that it is learning and will do better in the future?

Yes, the company cites adaptive/ feedback 'learning' as example of the 'deep learning' analytic, but the process is unclear how they will implement this feature in production.

Presumably, if the device sends you a false alarm, you just tag it as such in the app and over time it refines detection. But man, that seems like a major showstopper for the consumer market that just want it to work after they plug it in.

Read this IPVM report for free.

This article is part of IPVM's 6,743 reports, 909 tests and is only available to members. To get a one-time preview of our work, enter your work email to access the full article.

Already a member? Login here | Join now
Loading Related Reports