Avigilon Alta Appearance Search Tested
Avigilon Alta's cloud-based Appearance Search claims to "transform the manually-intensive examination of massive surveillance footage into accurate and useful results within minutes," but how well does it actually work?
IPVM tested Alta's re-identification-based appearance search (see guide) to see how it compares with competitive offerings, examining:
- How easy or hard is configuration?
- How does Appearance Search work in the Aware Mobile App?
- What search parameters are offered for Appearance Search?
- How well are subjects tracked across multiple cameras?
- How accurate are image-based search results?
- How accurate are color-based searches in full light, low light (~5 lux), and dark/IR scenes?
- How accurate is gender search?
Note: This test is the first in a series leading up to the Appearance Search Rankings Comparison, with Avigilon, BriefCam, and Verkada next in the series.
For Background, see:
Pricing *** ******** **********
**** ****** ******** *** ~$*** *** per ****** ******, *** ********** ****** is ******** ** *** ****** *******.
******** ** **-**** ******* **** ** Avigilon *** ********, **** ********** ****** offers ************* ***** ********, ******** * similar ****** ** ******** ** ******* (also ******** ** *** ********* ****):
Executive *******
***** **** ******* ***** **** **** suffers **** **** *********** ****** ********** its ******* ** ******* ********* ****** results ********** *** ************* *** **-**** Avigilon ***** ********** ******, *******, *** Alta ***** ********** ****** *** ********* faster **** * ****-***** ****** (******** reviewing *****).
Technical *******
**** ******* ***** **** ****'* ********** search *** **** ********* ** ******** subjects ****** ******** ******* (***** ** clothing *****) **** ******** ********* ********* per ******. *******, *** ******* ******** that ***** ******** ** **** ********** Search *** ****, **** ********** ******* being * ****** ***** **** ********* for ******** ** ***** (**** ** well-lit *********). ***********, ******** ********** **** occasionally ********, **** *** ***** ********** to *** ***** ******* ** ********.
************, ****** ****** ******* ******** ******, and ***** **** ************ ********** ** "female".
******, ***** ******** ******** ******* *** often ********** *******, *** ******** **** rarely ******* ****** ******** *******.
Appearance ****** ********* ** *** *** ******
***** *** ****** ************* **** ***** Appearance ****** ** **** *** *** and ****** ****:
Web ***:
Mobile ***:
Appearance ****** ** ** *******
**** ********** ****** ** ******* *** cameras ** *******, **** ** **** configuration ********.
**** * ****** ** ***** ** the **********, ** **** ****** ** an ********* ****** ***** "******," ********** of ******/******* ******:
Inconsistent ******** ****** ******** *******
****** *******, **** ***** **** ***** either ** ********* (*****/******) ** * picture ******** ******* ** * ******* across ******** ******* ** **********.
*** *******, ********* "******, ***** *****" returns ********* ** * ******* ****** several ******* ** *** **********:
*******, ********** ******* ******* *********. *** example, ********* *** "******, ***** *****" returns * ****** ** * ****** shirt, ***** *****, *** **** ****** multiple *******:
Gender ******* **** **** *************
***** ***** ***** ****** ***** *** alerts ***** ** ***** ** * reference *****, ****** *** *** ********* for **** *************:
**** *********:
****** ****** ** ***** ******* ** as * ****** ****** *********, *** reality ** **** *********** *** ****** of * ****** **** ** ***** is *** ***************. *** ****** **** neural ******** ** ******** ******** ************* for ******** ***** **** *** ***** female. **** ***** **** **** ****** could ** ********** ** **** ** as *******.
***** ** *** *******, **** **** of ********* ***** *** ****** **** “human ** *** ****” ***** ** review ******* **** ****** ** **** the ****** ** ********. ** ** not ********* *** *** ***** *** triggering ***** ***** ** ******.
Search ******* ** *** ***** *** *** *** ******
***** **** ****** * ********* ** 11 ********* ******, ***** *** **** choose ** ******/******* * ****** *****. By ********, ***** *******, **** ** Turing ******, ***** ***** ** ****** for ******** ****** ** * ****:
**** *********:
** *** ** **** *** **** experience ****** *** ***** ***** ********, and ** **** ********* ***** ***** in *** ******** ** ***** **** wish ** ****** *** *** ***** at * ****. ** ********, **** having ******* * ****** ** ********, our ***** **** ****** ********** ****** to ***** **** ******* ********.
Image ****** *** ******* *********
**** ***** ****** **** ***** ***** search, **** *** *** ********** ******* a ******** *****. *** *******, ** the ***** ******, **** ***** ******* 3 ******* *** * ********* *******:
******* ***** ***** **** ***** ***** search *** *** ***** ********** *** occasionally ******** **** *** ******* ** interest *** ** *** *****, ***** could ***** ****** ********* ****** ** miss ** *****:
**** ********* **** ** *** *** facial *********** ** ***** ******, *** it ** ** ***** *******:
**** **** *** *** ****** *********** in *** ******** ** ********, ******* that ** ********** ** ** ******** in ** ****.
Detection ** ******** ****** ******** ******* ********
**** ********** ****** *** *** ********* detect * ******* ****** ******** *******. In *** ******* *****, *** ******* is ******** ****** *** ******** ****** over * ****. *******, **** ******* provided *** **********:
Search ******* ********** ** ******** ** *******
**** ***** ********* ****** *******, ***** may ****** ** ****** **** ** accuracy ** ** *******:
****** *******, ********* ** ******** *** not ************ ******* *** **** ******** results *****. *** *******, * ** the ***** * ******* ** ********:
Search ** ***** ********* ********* ****
****** *******, **** ***** **** **** Aware's ***** ******** ** *** ****** was ********* ****, **** ******** ********** results ** * ****** ******.
*** *******, * ****** *** "***** shirt" ********* ******** ********** *******, **** only * ** *** *** ** results ******** ** "********" ********** * person ** * ***** *****:
*********, ********* *** "***** *****" ********* results ** ****** ** **** **** or **** ******:
************, ****** **** ***** ******** ** more **** *** *****. *** *******, an ****** ***** ** ******** ** "orange" *** ** **** ******** ** "red":
**** *********:
****** *** **** ***** ** ********, shadows, ***** ******* ** * ******, HDR *** *** ******* ** ** image *** *********** **** *** ******** color ** *** ******* **********. *** to ****, *** ****** ******** ** a ************ ********, ********* *** ******* as **** ** ** ** ********* it *** ** **, *** ***** showing **** ******* ** ****** * target ** *** ******. ** ***** intelligent *********, ******* *** ******** ** increase ** ********’* ******* ** **** a ****** ** ********.
Multi-Colored ******** ******** ** ******** ******
******** ******* **** **** *** ******** color **** ******** ** ******** ***** searches. *** *******, * ****** ******* a ***-*** ***** ******* ** ******** color ********:
Clothing ***** ********** ************ ********
***** **********, **** ***** ************ ******** results *** *** ***** ******* ** clothing. *** *******, **** "******, ***** shirt" ** ********, * ****** **** a ***** ***** *** ***** ***** are ********:
Objects ********** ** ***-*****
****** *******, **** ***** **** ******* were ************ ********** ** ********. *****, searching *** "***** *****" ******* * subject **** * ***** ***** ******* a ***** ******:
Colors ** ***-*****/** ****** ********** ******** ** *****
**** ****** ** ***** *** ***** (~5 ***) *** ** ****** **** detected ** *****, ** ***** *****:
**** ************, ****** ** ** ***** appear ** ******** *** "*****":
**** ***** ******* ******* ** *** parking *** ***** (~* ***):
****** * ******* *****, *** ******* may ** ******** ** ******** ******. Below, *** **** ******* ******* ** both "***** *****" *** "**** *****" searches:
Gender ****** ***** **********
**** ********* *** *****, ******* **** often ********:
*******, ******* ***** ******** **** ********* the **** ********** *** *******:
**** *********:
***** *** ** ************ ********** ********** for ******. ** ****** ** ****** networks, *** ****** ** ******* ** a **** ****** ** ****** ***** are ****** ** “****,” “******” ** “unknown.” *** ** ***** ** **** built, ****** *** **** ** ******** potential ***** *** *******. **** **** male *** ****** *** *** ******** attributes *** ** ** ******** *** a ****** ****** ** ** **** or *******. ** ********* **** ** operator ******* *** *******.
Events ***** ******* **** ********* ** ******
****** **** ***** ******* **** ****** was * ****** *********. *** *******, of *** ** ******* ******** *** "Person, ***** *****," **** * *** classified ** ****** "****" ** "******":
Only *** ******* ******** ** ****** **** ******** ********
** * ***** **** * ** more ********, **** * ****** *******'* attributes **** ********. *** *******, **** the ****** ******* ** ******** ** an ***** ********** **** * **** and ****** *******.
Versions ****:
*******:
- *** ****-***-*: ****************
- *** ****-****-**: ****************
- *** ****-*-**-**: ****************
- *** ****-*-***-**: ****************
*****:
- **** *****: *.*.*