Verkada First Generation Cameras Analytics Performance Decline
How do cloud video surveillance customers know if their provider quietly reduces the performance of their products? New IPVM testing shows this risk with Verkada.
Verkada claims their 2nd-generation cameras offer 5-7x longer people detection distance (5x lower PPF) than 1st-generation cameras, but IPVM verified that the detection distance of the 1st-generation camera is now significantly lower than when tested in 2020.
In this note, based on IPVM's testing and Verkada's response, we investigate Verkada's people detection PPF requirements and how it has changed over the years.
Executive *******
*** **** *** ******* ******* ****** the ********* ****** ********* ******** ** 20%, *** **** ******* ***** *** camera's *********** ******** ** **** **** 50%, **** ** **** **** ** just ** **** **** *** ******.
******* **** ******* ********* *** *** specifically ***** *** *** ********* *********** was ******* ******* *** *** *** actual *********** ******** ** *************.
********* ******** ******* ****** ** ******* do *** ***** ***** ** ***** and ****** *******, ***** ***** ***** throttling ** ****** *********** ******* *****' knowledge.
IPVM ******* ** ****** ********* ******** (*** ***********)
**** ******* ** ******* **** (***) **** *******'* **** camera ** *** **** *****, **** the **** ******** (~*** ***), *** same ************ ****** (~*'), *** *** camera ****-****, ***** * ******* *******. Over *** ***** **** ********* ** verify *** ********* ******** *** *** required *** ********** ****** *********.
Verkada ********
***** ******* *** *** ****** ************, they **** **** **** *** ************ with *** ************** ** ******* *** various *******:
** *** ** ** ************ **** our *** ************** ******* ** **** that ******* **** ***** **********, ********** capabilities, ****** ************/*********, *** ******** ****** the ******** ** ****** ********* ********.
******* ***** **** ****** ********* ******** differently ** ***** *** *** *** generation *******.
****** ********* ******** *********** ** *** camera ****** ***** ** ***** ***** resolutions *** ********** ************.
**** ** *** ** *******. *** question ** ***** *** *** ******** products' *********** ********.
Lower ****** ****** ********* ******** ** **** **** ****
****'* ******* ** *******'* ***-********** ******, CD41 *** ****, ***** **** ********** people ********* ******** ******* ** **** the ******** **** **** ** ****.** ****'* **** *******, ******* **** ************ ******** ****** at ~** **** (~** ***), ***********'* ******* ** ****** *** **** ****** ***** **** now ******* *** ******* ** ~** feet (~** ***).
** ****, ******* **** ******** ****** to ** ***** ** ***** ** the ****** *** ********** ********* (>** PPF ********) **** ** ****. *** video ******** ***** ** ****** ********* at ~** **** (~** ***).
Change ** ******* ********* **************
*************, *** ******* ****** ********* ***** shows **** ***** ** ******** *** consistent ****** *********. *******, *** ******** is ********* **** **' ** ****, to **' ** ****, ***** ** a **% ********* ******** *********.
Risk ** *********** ** ***** *********
***** **-**** ******* ***** ***** ** check *** ****** ********/*******, *** ****** to ******, ******* ****** ******** ******* automatically. ***** *** ****** ** / when ** ****** ***** ******* **** on-prem *******, ******* ******* *** ****** software ******* ***** *******.
********, ******* ******* ******* * **-**** warranty *** *******. **** *********** ** cameras **** *** **** **** ***** old ****** *****.
No ******** ******** *****
******* ************* ******* ******** ** *** direct-to-cloud ******* *** **** *** ******** firmware ********. ** ** *** ******** for ***** ** ****** ******** ** changes ****** **********.
*** * *** *** ** ***. I ** ******* ***** ** **** sort ** ********* *********** *** **** (I ***'* ****** ***). *** **** that *** **** *** *** **** results **** ***** ******* ** *** the ********** - ** ***** *** discussion ****** **: "**** **'** ******* this ***** ** ***** *******" ** actually **** **** *********** ** *****.
** ******, **** **** **** ****!
**** * ***** ** *** ***********: they ***** **** ******* *** ****** of ********** **** ** ***** ***** for *********. *** ******** ******** *** resolution ** * ****** ** ** analyzed ** ******** *** "*****" ** the ********* ***** ** *** *****. This ***** ****** ** ******* ** cloud ********* *****.
***** ***** ****** ** **** *** doing **** ** *** ** **** analysis ** *** ***** ****** **** on *** ******; ****** *** ***** to ******* **** ** ***** ********* processing ** ***** *****-*** ******* ** not.
[* **** *** * ***** ************ competitor ** **** **** *********** **** a ***** ** ****, ****** * don't ****** ** ** * *********, just * ******** ***********.]
****, ***** *** ******** ***** ******* may **** ****** *** *********** *********** in ***** *********. *** ***** ** the ***** ****** **** ******* **** Verkada ********** *** **** *** **** tests ******.
*.*. * **** **** *** * 'video ************ **********' *** ****** ***, I ** ****** **** ** * criticism. *'* **** ********** ********* **** honesty ** ******** * **** ******...
* "******** **************" ** **** **** V1's ** ****** ********* **** **** a ***** ******* **** ********* *** for ****** ********* *** ** *** data ******* **** *** ******** **** processing ***** **** ** ******** ** they ***** ******* **** ****** ** YOLO. *** *** ****** *********** "****" on ***** ******** *** **** ******** due ** * **** ** **********. It's * ***** ******** ** **** to ***** ** ******** ***** ****** for ***** ******** ** **** ***** not **** **** *** ****** ** they **** *** ******** ** **** them ***.
** ****** *********** ******* ** **** on *** ********** *********, *** ****** network ****, ********** ******* *** **** importantly *** ******** **** (***** *** be **** ********** ** **** ****).
**** *** *** ****** **** ** the *** ** *** ******* ***** its ***** ***** **** ** "******** AI." ***** ************ **** ** *** AI ********** *** *********** ** ** afterthought.
**** *** * *******.
**** ** *** **** ****** ***** used **** ***** ********* ** *** users ** *** *** ******.
******** ***** **** ** ***** **** how *** **** * ****** *** than ** ***** **** *** *********.