Neurala Brain Builder Tested (Train Your Own Analytics)

By Rob Kilpatrick, Published Nov 17, 2020, 08:53am EST

Do you want to train your analytics? For example, to detect a certain type of fish, tool, bird, buggy, etc.? While most analytics are pre-trained for people, vehicles, bikes, etc., Neurala says their Brain Builder app lets you train analytics for the specific objects you most care about.

But how well does this training work in real scenes?

IPVM Image

We tested the Brain Builder app on the new Security and Safety Things platform (tested here), examining:

  • Does it properly identify the objects it is taught?
  • Can it identify the object in a cluttered scene?
  • Can two objects be identified at the same time?
  • Can objects be identified at different angles?
  • How hard/easy is it to train?
  • How does it compare to Bosch AI Camera Trainer?

****: **** ** * test ** *******'* ***** functionality ***** ***** ********&** ***. ******* **** ****** an **-******* *** ***** version **** **** ******** and *********** ** ***** systems.

*******

***** ** *** *****, Neurala ***** ******* ********* identified ******** ******* **** it *** ****** ** close ****** *** **** uncluttered ***********. ******** *** simple, *********** **** ******** a "*****" ****** **** the ****** ******* ** the ***'* ********* ****.

*******, ** *** ******* limitations, **** ********* ******* in ******** ********* ******, when ******** ******* *** viewed **************, ** **** with ** (~*.** ***), or **** ******* *** carried ** ********. ************, objects **** ** ******* at ********* ****** *** proper *********** **** ********* points ** ****, ****** to ******* ***** *** error ** ** ** unclear *** **** ****** are ******** *** ******** recognition ** **** *****.

******* ** ****, ******* is ****** ********** *** general ******** ************, *** a ****** *** ** process **********, ******* **********, etc.

Neurala ********

******* ******* *** ********* feedback ** *** *******:

** ********* * ************* app *** *&** ** support * ***-**** ************* of * ****** *****-*** where *** ****** *** placed **** *** ****-******** produce ***** ******* ** order ** ****** *** need *** * **** to ******** ****** *** the ******* ****. ** with **** ** ****, certainly **-******** *****, *******, the ******** ** **** hand, ***. **** ****** performance.

************, ********* *********** ** cluttered ******, ******* ****:

**** *** ** ***** a ***** ***** **********, so ** **** **’* looking ** **** ***** of ********** ****** *** region ** ******** ** a *****. *** *** cluttered ***-**** ** ** successful, **’* ****** ********* a ********* ***** **** looks ** ****** ******** and/or ********* ***** ****** the ***** (**** ******** represented **** * ******** box ****** ****.)

Vs. ***** ****** *******

***** ******* ****** ******* similarities *********'* ** ****** *******, ** **** *** be ******* ** ********* specific ******* ** *** field. *******, ***** ****** Trainer ****** ** ******* false ********* ** ****** in ******* ******, ***** Neurala **** ***. ************, Camera ******* ****** ** longer ********* *** ** cluttered ****** ** ****, scenarios ***** ****** ****** detection ** ***** *******.

Neurala ******** *******

** ***** *******, ***** place *** ******* ****** in ***** *******'* ********* zone *** ***** *** "Train" ****** *** **** it. ** ***** ************* is *********, ****** ********** training *** ** ******** to ********* ********* ****** (discussed *****).

Multiple ****** *** ******* ********** ********

** ***** ** ********* objects ** ******* ******, Neurala **** ** ******* multiple *****. *** *******, below, *** ***** ** taught **** ** *** angle *** ****** **** pointed ****** *** ******.

***** ******** ** ********** angles, *** ***** ** recognized:

IPVM Image

Objects ********** ********

**** ******* **** ****** to ***** ******* **** were ********** ******** ***** when ********** ** *** scene.

Short *********** ****

******* ***** ********** ******* in *** ***** **** they **** ******* *****, present ** *** ***** for ~* ******.

Objects *** ********** ** ********* ******

***** ******* **** ********** recognized ** ****** **** relatively ****** ********** *********** or ***** *******, **** the ****** *** ***** at ********* ******, ** objects **** **********. *** example, *****, ******* ***** to ****** *** *** water ****** (***, ******) on * **** ****** with ***** *******.

IPVM Image

Object *** ********** ** ***** ****** ** ****

************, ** *****, **** realistic ****** ** ****, Brain ******* *** *** identify *******. *** *******, it *** ****** ** locate * ***** ** a **** **** ** other ***** ***** ** is ****** *** ****** on *** ****.

***** ******* **** ****** to ******** ******* ***** held ** * ****** walking ******* *** *****. Below, *** ******* ******* the **** ***** ***** was ********** ****** ** the *** ** ***** range, ~**', *** ** is *** **********.

IPVM Image

Objects ******** ** *** ***** (~***) ** ***** *****

****** **** **** ********** when **** ** ***** of *** ****** ** low ***** (~* ***), but **** ** ***** range.

No ********* ** **** **** ** (~*.****)

*******, *** **** ****** was ****** ** * dark ***** (~*.****) **** IR.

IPVM Image

No ************ ***********

** *** *****, **** one ****** ** * time ***** ** ********** in *** *****, ******* two ******* ***** ****** in ***** ******* **** picking *** ****** ** classify.

IPVM Image

S&ST *** ** ************ / **** ** ****

*** ***** ******* *&** app **** *** ********* to ***** *******. ** contains **** *** ****** monitoring **, **** ** settings ** ****** ****** displaying *** ******** ****** in *** *****-********* ******.

IPVM Image

S&ST *** *****

*** ** ** *** Neurala ***** ******* *** is ******, ********* ********* to ***** *&** **** by ********** ** ** the *** *****, **** pushing ** ** ******* cameras. **** *********, ***** may ****** *** **** UI ** ***** ******* or ****** ******.

S&ST *** **. **** *******

******* ******** ***** ******* as * **** ** the *&** *** *****, with * ******* ********** and ** **** *****.

*******, **** **** ***** a ****-******** ******* ** this ***********. ******* ******** that *** **** ******* is ********* ********** **** other *******:

********* *** ****** ******** by ***** ******* *** integrated **** ***** ************ that ******** *** ***** outputs ** ******* *** business ***** ** *** end ******** (*.*. ** trigger ** ***** ** a ******** *****, ** eject ********* ******** **** the ********** **** ** to ******* * ***** stick ** * ****-***** out *******.)

**** **** ***** * separate ********* *********** ***** can ** **** ** run ****** ******* ** Brain *******, *** **** not ******* ******** ********.

******* *** *** **** version ** ***** ******* starts ** $*,*** *** / **** ****. ********* starts ** $*** / year. **** ** ***** platforms *** ** *** on-premise, ** *** *****, or * ****** ** both.

Versions ****

*** ********* ******** **** used ****** *******.

  • ******* ***** *******: *.*.*

Comments (6)

Avatar

Morten Tor Nielsen

prescienta.com
11/17/20 05:49pm

************* ******** ***** ** ***'** into **** **** ** thing.

Undisclosed Integrator #1

In reply to Morten Tor Nielsen | 11/17/20 06:31pm

***... ******? **.**? ** Teachable ******* ******** ******** ****?

**** ***** **** **** the **** ****. (**** the ******** ********* ** erector ****--***** **** ****** cool.)

Joel White

11/18/20 11:38am

* ** ******* ** your ******* **** *** drill ******* **** ***** happen ** *** **** the **** ****** ********. Would *** ***** ***** be ******** ** ** Neuarala ********* *********** ******** to *** *** *** detector *** ******* **? In ********, **** *** train * ******** ***** can *** ******** ** moved ** ***** ******* equipped ******* ** **** each ****** **** ** be ******* *************?

Avatar

Rob Kilpatrick

IPVM | IPVMU Certified | In reply to Joel White | 11/19/20 05:33am

*********** ** *** ******** to *** ***** ** view **** ** *** trained **, *** ***** take ** ******* *** it ***** ***** ****** the ****** ** **** as *** **** ** within ~*' ** *** camera. **** ****** ** trained ************* ** **** other *** ****** ** moved ** ******* ** if *** **** ** object ******* ** ******* camera *** **** ** do *** ******** *****.

Avatar

jennifer shadbolt

11/24/20 08:00pm

*** **** ******** ***** analytics ** ********* ********* Cognitive ********?

John Honovich

IPVM | In reply to jennifer shadbolt | 11/24/20 10:08pm

********, ****** *** **** first *******. *** ******, here ** *** ******* page:********* ********—**** *** ** Developers | ********* *****

** **** ***. ** general, ** ***** ** testing ** ********* **** are **** *** *** users *** *********** ** deploy ******* *** *********** effort, ***** **** ** the **** ****** *** case.

**** ****, *****'* ********* or ***** **** **** Google ** ****** ***** be ****** *** ***** actually ********** ***** *** services.

Read this IPVM report for free.

This article is part of IPVM's 6,805 reports, 913 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