Verkada Trajectory Analysis And Enhanced Motion Search Tested
Verkada released its "motion search experience with trajectory analysis," saying their use of detection models and tracking algorithms outperform traditional approaches, but how well does it actually work?
IPVM tested Verkada's enhanced motion search and trajectory analysis and claims of increased people detection distance, and inside this report, we detail the full results of our testing, including:
- What is Verkada's "enhanced motion search"?
- What Verkada devices support trajectory analysis?
- How well does trajectory analysis capture people walking, running, or crawling?
- How well does trajectory analysis capture vehicles in motion?
- What are the PPF requirements for consistent people and vehicle detection?
- What / if any limitations were found with Verkada's trajectory analysis or enhanced motion search?
Verkada "******** ****** ******" ********
******* **** ***** *** ** ********* models *** ******** ********** *********** *********** approaches. *******'* ******-********** ******* ***** *** chips (********* **** *** ****) **** "provide **** ******* ******* *****" **** their ***-***** *************** (********* ***). ******* says ***** ****** ********* **** ** 10 ***, ******** *** ********** ********.
******* ****** ***** ****** ********* ** increased ** "*** **** ******* ******* zoom-in," ******** ** ** **** **** their ******** **********.
******* ********** ******** ** ************ *** *** ********** ******* *** its ******* ******.
**** ****** *** ******* ******** ** the ********* ****** ****** ******* ********** analysis: ****, ****/**/**/**, ****/**, ****-* *** CF81-E (** *********-**** ****).
****: ****** *** ******** *******, ** analyze *******'* *********** *********** ****** ***** on ****'* ******* *** ********.
Enabled ** *******
*******'* ********** ******** ** ******* ** default ** ********* (*** * *** CF81-E) *******. *******, ** ***** *****, no "*****" *** ********* *** * first-generation ******* ******* **** **** **** released **** *-* ***** ***:
Executive *******
** *** ******** ****, *******'* ********** analysis ****** **** ** ***-******* ***** on ****** *** ******** ** ******.
** *** ******** ****, ************ *** limited ** ********* ****** *** *** not **** **** ** ****-******* ***** due ** ************ ***** ******* ** dropping, ***** ****** *** ********* *********. Moreover, ************ **** ************ ******* ** people ******* ** ******** *********.
******* ************/ ********* ****'* ******** *** said **** *** ******* ** ******* this *** ********.
Trajectory ******** *** ****** ****** ********
** *** ***** *****, ** **** how ***** *** ****** ********** ******** on *******'* ******* ********* ** *******'* second-generation *******:
Comparison ** ******** **** ********
******** ********** **** ********, *******'* ************ *** ******* (*** free) *** ** *** ****** ***** to ****** ******* ***** ***** ************ as ********* ** ********, ******** *** practical *********.
*** *******, ********'* **** ******** ****** users ** ****** ** ****, *****, direction ********, ******* ****, *** **** (shown *****) **** *****-*********** ***** *** more, ***** **** ***** **** ********:
*** **** ** ******** **** ********,*** *** *******'* ********.
Trajectories **** ********* ** **********
******* ******** ************ **** ** ********** and **** *** ******* ******** ***** or ************ ** ***** ********, ****** it ********* *** ***** ** ****** the ************.
******* ******** ** **** **** **** don't **** ************ ********* ** **** video *** ******** *** ********* ** as * *******.
*****, ** ***’* **** **** *************, but ** ** * ******* **** we *** ********* ********. **’** ******** working ** ********* **** *******.
************, *** ******* *** **** *** list * ******** ** ******* ************ programmatically, ******** *** *** *** ***** motion ********.
****: ** **** ******, **** ********* the ********* ************ **** ***** ********* to ******* **** / ******** *** system's ***********.
No ****** ** *********
******* **** **** **** ***'* **** search **-********* *************, *** **** *** working ** ********* **** *******.
*****, ** ***’* **** **** *************, but ** ** * ******* **** we *** ********* ********. **’** ******** working ** ********* **** *******.
Tested *** ******** *** ********** ****** *********
*******'* ****** ********* ********** *** ******* **** ******** ** PPF (~***') *** ****** *********, *** we ***** *** ******* **** ******** ~12 *** (~***') *** ********** ****** detection *** ~** *** (~***') *** consistent ******* ********* ** ****-*** ******.
**** ******* ***** **** *******'* ***** CD41 **** ******** ~** *** *** consistent ****** ********* *** ~** *** for ********** ******* *********, ***** *** second-generation ******* ******* ** ********* ********* distance *** ****** *** ********.
*******, ** **** ********** **** ** on *** ******* **** (********** * cameras) ******** ****** *** *** ****** detection ** ****/** ****** **** ** well *** ******.
Inconsistent ********** ******** ** ******
***** ******* ******** ********** *** ****** of * ****** **** ******** ******* paths, *** ********** ******** *** ************, mixing *** ***** ** ******** ****.
******* ********* ** **** **** **** a ****** ** ******** ** ******* person, **** ****** ******* ** "**** or *** ************."
** **** ***-**-*** ********* (*.*. ***** doors, *******), **** *******’* ****** **********. That ****, ** * ****** ** heavily ******** ** ******* ******, ** is ******** **** ******* ***** **** or *** ************. **’** ********** ** tune *** ******* *** ******* ** that **** ** **** ****** ** happen ** *** ******.
Poor ****** ********* *** ********** ******** ***** *******
******* *** *** ****** ************ *** people ******* ** ******** ******, ******* their *****.
****, ******* *** *** ****** ****** running ** ****** ******, ** ** trajectories **** ********.
******* **** **** **** *** ********* of ****** ******* ******* ** ******* factors, *** **** *** ********* ***** tracking ********** *** ****** *** ****** releases.
*** ***** ********* ** ******** ****** depends ** **** *******, ********* ************ (shutter *****), ********, *** ********* ** movement. ** ******* *** ****** ** view *** **** ****** ** ******* the *********, ** ** **** *******, you’ll **** ** **** *** ****** in *** **** *** *** **** amount ** ****. **’** ********** ** improve *** ******** ********** *** ****** for ****** ********.
Solid ******** **** ****** ******
** *** *******, ******* ******** ***** trajectory ******** *** * ****** ****** walking ** *** ***** ** *** direction, *.*., (*******, ********, **** ** side, ***.):
************, ******* ************ ******* ****** ******** in *** *****.
False ****** **** ****** ********
***** ******* ****** "******* ******** *** *********** ** *********** people *** ********," ******* ******* *** ****** **** without *** ****** ******* ** *** scene ***** *** "******" ********* ***.
******* ********* ** **** **** ***** motion ****** *** ** *****, *** they *** ******* ** * ***.
****** *** ******** **** ** **. We’ve **** **** **** ***** **** up *** *** ******* ** * fix.
Trajectories *** ******** *** ********* ********
******* *** *** ******* ************ *** reversing ********, ******* ***** ******.
******* **** **** **** ****** *** short ************ *** ****-****** ********.
***, ** ************* ****** *** ***** trajectories *** *** **** ******* **********. If *** ***** **** *** ********* when ** ** ********, *** **** be **** ** *** *** ***** regardless ** *** ***** *** ********** is.
******* ***** **** **** ************ **** the ******* ** "*** **** ******** to *** **********."
** ***** **** *** ******* *** be *** **** ******** ** *** trajectory, ***** ***** ** ***** ************* hide **. **** *** ***** ******** over *** *****/********* **** *** **** is ********? **** ****** **** *** trajectory **********.
Solid ******** *** ******** ****** ********
******* ******** ***** ******** *** ******** moving ** * ******** ****.
Cameras *** ******** ******
******* ****: ** ** ****
******* ****-*: ** ** ****