Wyze AI Smart Vision / "Anything Recognition" Tested
Wyze has introduced "Smart Vision," which the company's VP of AI touts as being "anything recognition," a potential breakthrough in allowing users to pick exactly what they want to alert upon, but how well does it work?
IPVM tested Wyze's Smart Vision / "Anything Recognition," examining:
- How well does it detect objects that we custom-trained? We trained it to detect missing tools, open and closing doors, missing vehicles, different types of vehicles in a spot, etc.
- Can it detect missing or moved objects in a complex scene?
- How well does it detect trained objects outdoors or indoors?
- How simple is it to train?
- How commonly does it false alert once it is trained?
- What limitations does it have for checking the status of the object?
- How does it compare the Bosch Camera Trainer (see test) and Brain Builder app (see test)?
** *** ***** *****, ** *********** how ** ********* *** **** ***** Vision, ***** ** ****** *** ***** less **** * ******* ** ******, train *** ******.
Pricing *** *******
**** ***** ****** ** * ** additional **** ***** **** **** ****'* monthly ********** ************, ***** ***** $*.**/***** or $**.** *** **** *** ******. Smart ****** ** ********* ******* ** the**** ******* ********** ** *** ********* ** *** other **** *******.
**** **** **** ***** ****** ***** be ********* ** **** **** ****** models ***** **** ****.
Executive *******
**** ***** ****** *** ****** ********** with **** ** ***** ** ** a *********** ****** **** **** *** analytic **** ** **** ** ****** is ********, ***** ********* ****** ****** as *** ******** ****** *** *** been ********* *** (*.*., ****** *** bottle ** **** ********* ** **** the ****). **** ** ********** ** Wyze's ********* **-****** **** **** ******* notifications ***** **** ******* **** ****** events (*.*., * ****** ***** **** the *****, * ************ ** ****, then *** ****** ******* *** **** sanitizer ****** *** **-****** ******** ** finished).
** ***** **** ** ****** *** analytic *** ** **** ** ****** is ********, *** **** ******* **** it *** ******* ** ********, *** they *** **** ******* ** ************ on **** ********.
Wyze ******* ********: "***** **** ****"
****** *******, **** ********* **** ******* initially ********* *** ***********. **** ******* responded, ********* ** ** ****** ***** on **** **** ** ** ***** issues.
********* *** ********, **** ******* **** item ** **** ****, ******** ** 10 ****** *** ******* ****/**** *** increasing ** *** ****** *** ****/****. However, *** ******* **** ********** ********* and ***** ****.
Wyze's **** ******** ** ****** ***** ****** *******
****'* *********** **** ********* ** ****'* test ******** ** ******* *** *********** of ***** ******, **** *** **** response ****** *****:
**** ********* *** ******* *********** ****** on ****** ****** *** *** ****** of ** ****** ** ** *******, which **** ***** ** ** * fundamental ****** ****, *** ***** **** is ******* ** ************.
** *** ******* ** ******** *** optimal ******* ******* *****, ***********, *** cost, *** ***** ****** ****** ********* primarily ****** ** ****** ****** ** activate. *******, ** *** ******** ******* on *********** ********** ********, **** ** rule-based *** ****-***** ********, ** **** as ********** ********** *********.
**** ****** *** ****** ******** *** training **** ********* ** *** ********** of * ******* "*****," *******, ** our *******, **** ***** ********** ******** did *** ******* ***********.
*** ****** ** ****** ******** *** effective ******** ** *** ***** ****** depending ** *** ********** ** *** task ** ****. *** **** ******* tasks **** **** ****** *********, * smaller ****** ** ****** *** *******. However, ** *** ********** ** *** task *********, ***** **** *** ********* of ********* *** *** ******* ******* of ******, *** ***** ***** ****** may **** ********.
******* **** ******* *** ******* ** "experimental," ***** ****** ** * ***** released ******* *** ********* **** * Cam ******* ************ ** *** **** v2 *** ** ******, **** **** camera ****** ** ******* ***** ****** later **** ****, *** *** *******.
***********, ** *** ********** ******* ** implementing * ********* ** *********** ******* the *****'* ***********, ****** ** **** it ** ********** ** ********. **** we *** ********* **** ************ ******* to **** **** ****** ****** ** the *** **** ** ****. *** goal ** ** ******** **** ************ and ****** * ******** **** **********. Stay ***** **** **!
**** *****, ** **** ******* ** still ** *** ***** ******, ** highly ***** *** ******** *** *** provide. ** *** ***** ** ****** in **** **-***** *********** **** ***. Please **** **** ** ***** *** to ** **** *** ********** ********* or ******** *** *** ****.
***** *** *** **** ********* ******** and *******.
Worse **** ***** ****** *******
** *** *******, *** **** ***** Vision ********* ***** ********* ****** ******* (*** ****). **** ***** ****** *** ****** with *****/********** ****** **** ****** ******** anywhere ** *** ***** *** **** checked *** ******'* ****** **** ****** was ********. ** ********, ***** **** constantly ****/****** *** ****** ** ** item ** *** ***** *** ****** count ** ******** ***** **** ** is *******.
*******, ***** ****** ******* *** ****** with ****** ** *** ***** ***** identified ** *** ****** ** *** trained **, ***** **** ******** ***** alerts **** ****** ****** ****** ******.
***** *** **** ** *** ***** Camera ******* **** ** **** ****, we ********* **** ***** **** ***** had **** ** ******* ***** *** test.
** *******: ***** *** *** **** any ****** ** *** ****** ******* since *** **** ****** **.
Worse **** ******* ***** *******
******** ********* ***** ******* (*** ****), **** ***** ****** ********* *****. Brain ******* *** *** ******* ***** alerts ** ****** ** *** *****, while **** ***** ****** ********** ******** false ****** ** **** ******.
Motion ********* *** ********* ****-**** ****** ******
** *** *******, **** ***** ****** had ****** *** ** * *********** design **** **** ****** ******** ****** on ****** ******** *** *********** *** to ********* *** ****** ** **** as ****** ** ********, *** ** the ****** ****** ****** ** *** trained *****(*.*., * **** ******/******).
**** ***** ** ********** ** * mandatory ****-**** **** ** ~** ******* that ****** ****** ****** ** ***** alerts ****** *******.
** ***** ** *** ***** *****, motion ****** *** **** ****** ** show ******, ******* *** ******* ******* the ****. ****, *** ******* ****** through *** *********, *** *** ****** is ********* ******* ** *** ********* 15-second ****-**** ******. ******, *** **** status ***** **** ******* *** ******* closing *** **** *** ********** *** of *** ***.
**********in ******* ******
** * ******* ***** **** ** a **** ** *****, **** ******** to ****** **** ** **** ** missing, **** *** ********* ******** * missing **** **** *** ******* ******** (the **** **** ****** ** ********).
*******, ** *** ****** ********** ****** testing ** *** ****** *** ********* incorrect ** *** **** *** ******* before *** ****-**** ****** ******* ******* the ****** ** ** *********. ******, if ***** *** *********, ****** **** be **** ******* *** ***** *** missing.
Can't ************* ******* *******-***** ********
** *** *******, **** ***** ****** was ******* ** ***** ** * specific ******* ******* * ******* ****, it *** ********** ** ***** ****** would ****** **/**** * ******* ******* was ** *** ******* ********, ********* false/incorrect ******.
** *** ******** ****, ** * motorcycle ** ******* ******* *** ****** in *** ******* ********, ***** ****** correctly *******/********* *** ******* ******* *** missing.
Constant/Nuisance ****** ** ***** **** ******** ******
**** ******** *** ******* ** *** presence ** * *******, **** ***** Vision **** ********** ***** *** ******** vehicle ***** ** *** ***** *** to ****** ********** *** ******** ** run.
*** *******, ******** ** *** ********** of *** ******* *** ****** *** alerts ** ******** ** ** ****, despite *** ****** ****** ** *** trained ******** ********* ********* (*.*., *** present/gone).
Generally ******** ** **** ****** ******
**** *****/******** ** ****** *******/****** **** infrequent ******/***** ** *** *******, *** results **** ********* ******** **** *** objects' ****** *** ******* ** ******* (outside *** ******* ****-**** ****).
Versions ****
*** ********* ******** **** **** ****** testing.
- **** **: *.**.**.****