Avigilon Unity Appearance Search Tested
Avigilon was the first manufacturer to offer Appearance Search in 2016, but how does it work in 2023, and how does it compare to cloud-based offerings such as Alta or Verkada?
IPVM bought and tested Avigilon Unity Appearance Search to see how it compares to offerings such as the previously tested Avigilon Alta and Verkada, examining:
- How does Avigilon Appearance Search compare to cloud-based offerings Alta and Verkada?
- How does this compare to Avigilon's facial recognition?
- What are the tested key strengths and weaknesses found?
- What search parameters are offered for Avigilon Unity, and how well do they work?
- How well are subjects tracked across multiple cameras?
- How well are image-based search results?
- How easy or hard is configuration?
- How well are color-based searches in full light, low light (~5 lux), and dark/IR scenes?
- How well do gender and age-based searches?
- What if any issues were found during testing?
Note: This test is the 3rd in a series leading up to the Appearance Search Rankings Comparison, with the first two tests being Avigilon Alta and Verkada, with i-PRO Active Guard and BriefCam coming next.
For Background, see:
Features **********
*** ***** ***** ******** *** ******** offered **** ** ******:
Pricing **********
******** ** *** **** *********, **** Avigilon ********** ****** **** *********, ** it ** ******** ** **** ********'* highest-cost ********** ******* (~$*** **** *** camera). **** *** ******* *** *** least *********, ********* ** *** **** annual ******** ******* (************ $*** *** $199 **** ********).
****** *********** ** *** ******** *** Avigilon *****'* ********** ****** *** ** a ******** ******* ******** ******* ~$***+ MSRP *** ******.
** *** ******** *******, ** ****** our **** ********.
Evaluating ******* / ***********
** ********* ********'* *********** **:
- ***** *** ****** ***** ** *** many ***** * ******* *** ******** on ******** *******.
- ****** *** **** ******* ***** ********'* appearance ******.
- ********* *** **** ** *** ******* shown **** ** **** *******, *** many **** ** ****** ****, *** how **** ***** *** ******* ******** was *** ******** ** *** *******.
***** ** ** ******* ******* * searched ***** (** *** ****, ***** in ****) - *** ** *** 11 ******** ******* *****, ** *** the *******, * ** *** *** subject *** ***** **** ** ********* where *** ******* *** ******** *** was *** ********:
** ******* ** **** ******* ** there **** ********* **** * ******* was ******** ** ******** ** **** time ******, ** ** *** *********. So ** ******** ******** *** *** instances ** *** ***** *******, *** accuracy ***** ** ********** ** **% (i.e., *** / ***).
Executive *******
** **** * ***** ** *******, Avigilon ********** ****** *********** ******** ********* based ** ******* * ***** **** and **** **** ********.
***** *** *** *** ******* ***** based ** *** *******:
- **%+ ******** **** **** *** **** were *****, ***** ** *** ****** (~50+PPF).
- searched ** ******* **** *** *** have * ***** ****, ******* ***** to *** ****** (~**+***) **** *******.
- far **** *** ****** (~** ***) or **** ** **** *** *******, motion **** ******** ****** ** ******** subjects ****** ******** *******.
***** ** ** ******* ******* ***** the *********** ** *** ********** ****** was **%+ ******** **** *** **** and **** **** ***** ****** ***** the ******** *** is ** **** *** *******:
*** *********** ******* ****** **** ********'* algorithm *** ********* **** ******* ********:
- * ***** ***** ** * ******* face *** *** **** ****** ********* on ********.
- ******** ***** *** * ********* ******. Without * ***** ****, ** ***** more ****** ** ******** ******** **** similar ******** ******.
- *** **** / ***** ** ** individual *** * ******** ****** ** accuracy; ****, *** ****** **** ******* builds **** ** **** ****** ** be ******* **** * ****** ** a ******* ** * ******* *****.
******** ********* **** **** ********* **** exist, ****** *** *** ******** *****-***** details.
Avigilon ***** ****-******* ****** **. ******** ****** ***********
***** ********’* ****** *********** ********* ************* fewer ***** ********* (*.*., ******* ** people *** **** *** *** ******* searched ***) **** ********** ****** ***** face ********, *** ****** ******* **** true ******* *** ** *** **** expensive **** ******** ****** ***********. ***** face ********, ********** ****** ******* **** more ******* **** **** *** ******* is *** **** *** ****** (*.*., lower ***) *** ***** ******* ***** quality, ***** *** ** ******* **** trying ** **** * *******.
*******, ****-*******-***** ********** ****** **** ******* far **** ******* ** ****** *** are ********* *** *** ******* *** is ********* *** *** *** **** more **** ** ** *** *************. For ***** *** ******* ** ******* facial *********** ******, ****-*******-***** ********** ****** is *** **** ********* **** ***-** Avigilon ****** ***********, ***** **** ** $500+ *** ******, * *********** *********** cost.
Avigilon ***** ********** ****** **. **** *** *******
*** ******* ***** ***** *** ******* that ****** ** ******** ***** ****** cloud-based **** ** *******:
- ***** *** **** ******** ***** **********, e.g., (***** ****/****, ****-***** *****, ~**+***), Avigilon ***** *** **%+ ********, ******* Alta *** only ******** ******-**-***** ** * ****** recognition *******.
- *** *****-***** ******* *** "****-***-****" *** require ***** ***** (** **** ******* by *******), ******* ******** ***** ******** more ***** ** ********* *** ****** scene ********, ***** **** ** ** 24 ***** ** *** *******.
- *** *****-***** ********** ******** *** "****" features, ******** ** *** ********* ***** for ****, ***** **** ~$*** **** per ****/*** ****** **. ~$*** **** per ****/*** ****** *** ******* **. ~$300 *** ****** (***-**** ****) *** Avigilon ***** *********.
*******, ********, ****, *** ****** *** performed ****** **** ****** ** *** / ******. **** **** ** ***** rankings ** *** / ****** ********* in ****.
Avigilon's "**** *****" *********
**** ***** **** ********'* "**** *****" hierarchy ******* ******* ********** ***** * person ** ******** (*.*., ****, ******, etc.) *** *********** *****, ***/****** *****, followed ** **** *****, ******, ***, etc.
** *** *******, ********* ** ***** facial/body ******** ******** **%+ ******** ** full-light ******, ******* ** *** ******* was *** **** *** ****** (~*****), accuracy *** accuracy, ** *** ******* ** ****** by ****** **** ***** ****** *** body *******, *** ******** **** ***** to *** ****** (*.*., ~*****+)
******, "**** *****" ********* ***** ********* (i.e., ****** ******* *** ******* ******** for), ** ** ********* *** ******** people **** ***** *** ** *** search ******** ((****, ***/******, ******, ***.)., showing *** *** ~*** ******* ** a ******, ***** *** ** ******* refined ** ********* **** **********.
******** ********* **** **** ********* **** exist, ****** *** *** ******** *****-***** details.
Appearance ****** ********
***** ** ** ******** ** ******** Appearance ******:
Some ******** ******* ************ **** ********** ******
**** ******* ***** *** ** ********* were *** ********** **** ********** ******.
******** ******** *** ********* ******* ********* (current) *********** *******:
*** ******* *******, ********** ****** ** not ********* ** *** **** **** of *******, ** **** *****, *** cameras, ** ******* *******, ** *** (license ***** *******) *******, ** **** H5 *** *******
Additional ******** ******** *** *** ***** *******
**** *********,** ***,** ********* *, ** ******** ******** ** **** ** *** *** party ******* **** ********** ****** ** Avigilon ******* ******. *******, ***********, **** BriefCam, *** ****** ************* ********, ******* the **** *** ********** ********.
Bug ***** ** ***** ******
***** *** ****** ******** ******, **** found * *** ** *** ***** upload. **** ********* **** ****/**** ******, the ****/**** ******** **** *** ********** displayed:
******** ******** *** ********* ******* ********* the ***** ****** ***:
*** **** ** ******** **** *** image *********, ******* ** *** ******* on * ********** ** * ***** issue ******* ** **** *** ******* where **** ****** **** * ***** aspect ***** (**** ** **** ***) display ** ********* ********* ** *** face ******* *****. **** **** *** impact ******* *******.
Search ** ***** **** ****/**** *******
******** ********** ****** ** ****/**** ******** 90%+ ******** ******* ** **** ***** scenes **** ********* ** * ***** face/body ******* ** *** ******* *** subjects **** ***** ** *** ****** (~50PPF+)
** *** ******* *****, *** ******* is ******** ****** * **** *** several ******** *******, **** ** ** 20 *** **** "********" ******* ********* the *******:
******** **** **** :
***** ** ***** ***** **** * face *** * **** **** ******* can ** ******** **** ******* *** most ************* *** ** *******, ******** the ****** ** ******** ** ******* the **** ******** ** *** ***** being ******** ** ****’* ** *** uploaded *****.
Search ** ****,
**** ***** **** ********* *** * subject ***** ** **** * **** produced did *** ****** ***** **** *******, and *** ****** ********* ***** ***-******** faces ** *** *******.
** *** ******* *****, ***** * face ******, ********* ** *** ******* with * ******** **** *** ********** populated *****. *******, ***** * ****/**** profile **** *** **** **** ****** reveals ********** ********* ** *** ******* where * **** ** *** ********:
** ** **** **** **** ********** using ********** ******, **** ***** ** an *****, ** *** *** *** be ****** ** ***** ******* ******** before **** ******* **** ********* *** only * ****.
"Starred" ******* ******** ******* ******** ********
**** ******* ***** **** ****** * "star" ** *** ****** ** ******** increased ******* ******** ****** *******, **** the **** ******** ******* ***** * face/body *******, ******* ** ******** ***** results.
** *** ******* *****, ********* * subject **** **** * **** *** body ******* ********** ******** *** ******* of *** ******* *** *** **** period, **** ******* ******** *******:
*******, *********/********* ** ***** **** ** face ******* ******** ********** *******, **** subjects ** ******* ******** ********* ****** than *** ******** *******:
******** **** **** **** ***** *** used *** ********* ** ******** ****** longer ******* ** ****, ***** *** full **** ******* ****** ******* ** be **** **** *** ******* ** facing **** **** *** ******:
*** *** ** ******** ***** ****** an ********** ** ******** ** ** found **** ****** ****** ******* ** time (***** **** *** **** ******* clothing) ** **** ********* ***** * photo ***** ** ******* *** *** example. ** *** **** ******* ****** accurate ******* ** **** ****.
*** *** ** * **** **** profile *******, ****** ******* ** ** made *** ********* ** *** ********** of ******** **** **** *** *** be ****** *** ******. ********, ** a **** ‘*****’ ** ********* ********** results ** ***** *******, *** ****** can *** **** ******** *** *** multiple ******** ** **** **** **** images **** **** ********, ** **** body ****** ****, ** ******* * more ************* *** ******** *** ** results.
Images **** ******** ********/** ******* ********
**** ******* ***** **** ****** **** multiple ******** ** ** ******* **** rejected *** ***** ******:
Confidence **** ********, ******** *** *******
******** ******** ********** ****** *** *******, and ***** *** ********** ** "**** results" ** ******** ** ****** ******* in **** ********** ******** ** *** search ****** **** ** ******** *****.
******** ******** *** ********* ********:
*** ****** ** *** **** ** the ***** ** *** ****** ** the **** ********* *** ******** (** other ******* ** *** **** ****** query) *** *** ***** ** ** absolute ***** *****. ** **** **** indicate **** ***** ******* ** **** contain *** **** *******, ******** ** other **** ******* ****** *** **** overall ****** **** ******. ** **’* possible ** **** * ****** *** if ****’* *** **** ****** *** a ***** *****, **** ** **’* not * ****** ******* ***** *** the ***** *******.
*******, **** ***** **** ******** ********** Search ******** ********** **** **** ** associated *******, ***** **** ********* ******. In *** ******* *****, ******* ********** bars *** *******, ***** ***** *** no ******* *****:
**** ***** ******** *** **** *** occurring. *******, **** *** *** ******* a ********, ** **** ****** *** report ** / **** **** *******.
Camera ******** ******** *** ******** *******
******** ********** ****** ******** ***** ** configure *** ***** *** ***** ****-******** for ******** ******* - ** ********, Alta *** ******* ******** ***** *****.
******* *** ***** ******** *** ****-******** applied *** *** ******'* ***, ********** results ** ********** ******** **** ******, including ********* ******* ********* ** ******** results.
** *** ******* *****, *** *** was ******* ***** ****-******** *** ********, and * **** *** ******** ** a ****** *** "****" **** ** person ** *** ***:
Color ****** *******
****** *******, **** ***** **** ********'* Color ****** ********* *** **** ******** results ****** ***** *** ********** **** accurately ********** ****** ********.
** *** ***** *******, ********* ** hour ********** ******** ****-********** **** ******* up ******** ****** ** * ******* in * **** *****:
Color ****** ** *** ***** (~* ***) ******
**** ******* ***** **** *****-****** ******** accurate ******* ** *** ***** (~* lux) ******. ** *** ***** *******, a ****** *** "******" ******** ****-********** results ** ~* ***:
IR ****** **** ********** ***** *****, *****, ** ****
****** ** ** **** **** ********** by ***** ***** *****, *****, ** gray. ** *** ******** *****, * search *** "***** ***" ******* ******** IR ******, ***** * ****** *** "orange ***" ******* ** ******* *** the **** **** ******.
Inaccurate ****** ******
****** ****** ********* ******** ********** *******, with ********** **** *** ************ ********** with ******** *******.
** *** ******* *****, * ****** for "******" ********* **** ******* ********** males **** **** ********** ******* ******* being ******* ** **** ******.
Issues **** **** ***** ******
**** ***** **** *** **** ***** search *** *** ********, **** ********* making **** ****** ****** ** **** scenes.
*****, * ****** *** "******" **** generates ******* *** ******** **** ***** hair, **** **** *** ******* *********:
********* *** ***** **** ********* **** results, **** ****** ******** ********* ** the "***** ****" ******:
******** ******** *** ********* *******:
******** *** ****** *** * ***** appears ** *** *****, **** **** if ***** *** ******* ** **** lighting, ***** **** *** ******** **** closer ** *****, ** ***** **** may **** ****** ** *****, *** example, *** **** **** ****** *** accuracy ** *** ****** *******
Issues **** *** ******
**** ******* ***** **** *** *** search *** **********. ** ****** **** both ******** *** ****** *******, ********** Search ******** ****** ** *** "*****" search **** *** **** ********** ** children:
******** ******** *** ********* *******:
******* ****** ** *****: *** ******* are ****** ** ********** ***** ** match *****, *** *** **** ******* are ****** ******** ** **** ** there *** ******* (****** ** ********) detected ****** *** **** ** *** search *****, **** ** **** ***’* score ****** *** *** ******** ****** query.
Versions ****
*******:
- ******** **.**-****-***-**: *.**.*.**
- ******** *.**-***-***-**: *.**.*.**
- ******** *.**-***-***-**: *.**.*.**
VMS *******:
******** ******* ******: *.**.**.*
*** ****,
****** *** *** *******. ******** *** report *********** ** ******* ***** ** well:
**** ***,** ********* *, ** ******** ****.
*******, ***********, **** ********, *** ****** manufacturers ********.
********, *** ** ******** *** ****** manufacture ********?
** *******, ******** **** *** **** additional **, ******* ******** (***** ****) requires ********** ** *** *** ***** cameras.
** *******, *** ** ********* ** used **** **** *** ***** *** servers **** *** *******-*****. **** ** the ***** ******** *** *** ******* about.
** *** *** *** *****/***** **** have ******* *** ********** ***** ** the **** *** ** ** ***** hardware ** ****** **** *** ***** cameras. **** ** ********.
****, **** ** ********* ***. **** also *** * ******** **** ****** just **** *****
****:
*****:
**** ** *********. ***** *** ******** solutions **** ******** *** **** ******* adding ********* ******, **** ***, **** to *** ***** ******* *** ******** cameras **** ***'* ******* ** ***** in.