i-PRO Active Guard People Attribute Search Tested
i-PRO (formerly Panasonic) touts Active Guard as having " the industry's highest number of search attributes" but how does it work, and how does it compare with other appearance search systems such as Avigilon?
IPVM tested i-PRO Active Guard appearance search, examining:
- How well do the various search parameters offered for i-Pro Active Guard work?
- How accurately are subjects tracked across multiple cameras?
- How accurate are color-based searches in full light, low light (~5 lux), and dark/IR scenes?
- How accurate is gender search?
- How accurate are additional search parameters, including bag/no bag, age, direction, sunglasses/no sunglasses, and mask detection?
- How easy or hard is configuration?
- What if any issues were found during testing?
- How does i-PRO Active Guard compare to Avigilon Unity Appearance Search?
Note this is the fourth in a series leading up to the Appearance Search Rankings, with the first three tests being Avigilon Alta and Verkada, Avigilon Unity, and Briefcam coming next.
For background on the technology, see IPVM's Appearance Search Guide.
Features **********
*-*** ****** ***** ** ******** **** the ******** ***-*** * ** *-****** *******(*** ** ******-***, **** ********* ******) and ****** **** ****** ********** **** Avigilon *** *******:
***** *-*** ****** "**** **** ** people ****** ******* *** **********," **** testing ***** **** **** ******** **** color ** * ******** ******/*********:
*******, **** ********* *** ***** ********, i-Pro *** **** ****** **********.
Facial ***********/******** ******* **** *******
***** ****** *********** ** ********* ** no ********** **** **** *-*** ****** Guard, **** ******* ****** *********** **** this ****. **** **** **** *-***'* facial *********** ** * ******** ******.
**** **** ******* ******** ** **** report, ***** **** ** ******** ** upcoming *******.
Testing ***********
**** ****** *** ******** ** *-*** Active ***** ***** *** ********* ***********:
- ***** *** ****** ***** ** ***** subjects ******** ** *-*** *******
- ***** *** ****** ** ***** * subject *** ******** ** *-*** ******* in * ***** **** ******
- ********* *** **** ** *** ******* shown **** ** **** *******, *** many **** ** ****** ****, *** how **** ***** *** ******* ******** was *** ******** ** *** *******.
******** *** ******** ***** ** *** number ** ***** *** ********* ******** subject *** ******** ** * ****** vs. *** ****** ** ***** ** incorrectly ******** ******* *** ******** **. the ****** ** ***** *** ******* was *** ***** ** *** ****** but ******** *** ***** *********:
** ******* ** **** ******* ** there **** ********* **** * ******* was ******** ** *-*** ** **** time ******, ** ** *** *********. So ** *-*** ******** *** *** instances ** *** ***** *******, *** accuracy ***** ** ********** ** **% (i.e., *** / ***).
Executive *******
** *** *******, *-*** ****** ***** accuracy ****** **** ****** ** ******** depending ** ***** *******: *** *****, obtaining * ***** **** ** *** subject, *** ***** *** "******* *******" or "*******" *-*** ********.
*** ******* ********, ** ~**%, *** achieved **** *** *** ***** *** 100+, *** *******'* **** *** ******* captured ** *** ******, *** *** "exactly *******" ******* *** ****.
*** ****** ********, ** ~*%, *** achieved **** *** *** ***** *** less **** ***, *** *******'* **** was *** ******* ****** *** ******, and "*******" ******* *** ****.
***********, *-*** ****** ***** ************* *** complex *** ****-*********, ********* ***** ******** software ********* *** *** ******* ** the ******* ********* *** ****** ** install ****** *****.
i-PRO ****** ***** ******** ** ******** *****
******** ***** **** ********** ******* ******** across ******* *** ******** ******* *** to ********'* ********* ****** ** ********** Search **. *-***'* *-* ********* ***** for ********* ********.
***** ******** ********** ****** ******* ~**% accurate ******* ***** * ***** ****/**** image, *-*** ******* ~**% ******** ******* but ******** ****** *** *** ****** for ******** ***** *** ******** *******, ~50PPF **. ~******.
*******, ***** ***** ***, ****** ****** yielded ****, both *******.
i-PRO ****** ***** ********
***** ** ** ******** ** *-*** Active *****:
"Exactly ******* ***************" ****** ******** **** ***** ****/****
********* ***** ****** **** * ***** face/body ******* ~**% ******** ******* ***** the "******* ******* ***************" ****** ** i-Pro ****** *****.
**** ********* ***** "******* ******* ***************," Active ***** ******** *** **** **** attributes (***/****** ******** ****/*****) *** **** attributes (******, ***, **********, *****, ***.). During *******, ** ***** **** ********** not ******** **** "**** **********" *** omitted **** *** "******* ******* ***************" search.
** *** ******* *****, ********* *** a ******* **** * ***** ****/**** (>100PPF) ******* ** ******** *******:
*******, **** ********* *** "*** ***," across *** **** **** ********** ******* there **** ** **** ********* ** the ******* **** ****** ******* *** omitted **** *** "******* *******" ******:
*** ******** ******** ***** "******* *******" characteristics ****** ******** *******, *** ******* must ** *** ** ** ******* both *****/****** ** ****** ******* **** clearly.
"Similar ******" **** ********, ********** ** ******* ********
** * ****** *** "******* ***************," fewer ********** *** ********, ** ****/**** characteristics *** *******, *****, ****** *******, we ***** ********* **** ******* (****** detections), ****** **** **** ********.
** *** ******* *****, * ******* with * ***** ****/**** ** **** in *** "******* ***************" ******. ***** the ****** ********** ******* ** *******, it **** ******* *** ********* ** other ******** (*** *** ***** ** shown):
*-*** ******** *** ********* ************* ********* the ********** ******* "******* *******" *** "similar" ********:
*** *************** **** *** ********* ****:- Search *** ******* ******* *************** (**** false *******)’ = ****** **** ****/**** characteristics *** ***** **** **********% ** body ***************- ‘****** *** ******* *************** (less ******* *******)’ = ****** ******* head/face ***************, **** **** **** **********% of **** ***************.
Poor ********* ** ******** ****** ******** *******
******* ****** ******** ************ ******** * subject ****** ******** ******* *** ** i-PRO's *-* ******** *********. *** *******, if *** *** ***** *** ********** at * **** ********** *****, ********* of *** ******* ** ****** ****** were ******* ***** *** ******* *** changed ******* ** ** ******** ******* confidence ****** ******** ** *** ******** parameter.
Watchlists ************ *** **********/*****
*-*** ****** ***** ******* ***** ** configure ********** *** **** ********** *** faces, ** ***** *****:
*******, *********-**** ********** *** ******* *********, as *** ******** ******** ** **** detection ** * ****** ******* *** impacted ** ******* *** ******* *** facing ******/**** **** *** ******, ********, and ******** **** *** ******.
** *** ******* *****, *** **** subject ** ******** ******* ****** *** away **** *** ******. *******, *** top ***** ** ******** ** "*****" in *** **** ***** *** "*****" in *** ***** *****. ***********, *** to *** *******'* *****/********* ** *** camera, **** ********** *** ******** ** the **** ***** **** ** *** right:
Color ****** **% ******** ****** ****** ****** "*****"
**** ******* ***** **** ***** ****** was ~**% ******** ** ****-***** ******, enabling ***** ** ****** *** *** or ******** ********* ****** *** ***/******. However, *** ********** ***** *** ************ detected ** ***, ******, ** **** color ****** *******, ~*% ** *** time.
** *** ******* *****, ****** ***** returns ******** ******* *** "*****" ** "orange" ***:
*******, ** ********* *****, *** ********** color *** ************ ******** ** "*** color" (as ***** *****, * ****** ***** is ******** ** * ****** ***:
*-*** ********* ***** ********:
*** ********** ***** *** ****** *** results.
*** ***** ***** *** ** ******* from ******** ** ***** ******, ********** returning ****** ** ******** ** ****/***** shirts ********** ** *** **********, **** accuracy ** *** ** *%. ** the ******* *****, ** ******* ******** is ******* * ***** ***:
*-*** *** *** ******* ** *** question ***** ******** **** ********* *** green ******.
Inanimate ******* ******** ** "******"
**** ******* ***** **** *-*** ****** Guard ******** ********* ******* ** ******** as ****** *-**% ** *** ****.
** *** ******* *****, ******* ****** of ********* ******* *** ******** ** a ****** *** "******" **** * hours:
*-*** *********** ****** ******* ***** ****** areas ** *** *** ** ******/******* detections ** ********* *******:
** ********* ***** *** **** **** setting ***** **** **** ******* ***** from ***** ******** ** *** *******.
IR ****** **** ********** ** "***** *** ***** *****"
****** ** ** ***** **** ** searched ***** "***** *** ***** ***** only," *** **** ******* **** "*******" or "******* *******" ***************-***** ******** *** color-based ********.
*******, ****** ** *** ***** (~* lux) **** ******** ** ***** *** white ***** ******, ****** **** *** negligible ** **** **** ***** * occurrences **** ~* ***** ** *******.
********* ******* **** *** **** ****** to ******** ** "***** *** *****," IR ****** **** ** *****. **** ~3 ***** ** *******, **** **% of *** ****** ******** ***** "***** and *****" ***** ********* ********* ******* with ** ****** **. all ******.
Results ******* **** ****** ****** *** ** **** ********* ***********
**** ******* ***** **** ***** *-*** Active ***** *** **** ********** ******* for ******, **** ******* **** ******* from ******-***** ********, ** *-*** ****** Guard ******** **** ********* ** ********* gender.
** *** ******* *****, **** ** of *** ****** ******** ****** *** same ****** *** ********** ** ****** "male" ** "******:"
*-*** *********:
*** ****** ****** **** ****** ********. If *** **** ** *** ********, then ********** **** ** ******, ***, sunglasses, **** ****, *** ***** **** not ****** *******.
Accessory-Based ******** **********
****** ***** *** ******* ********** *** accessory ********, ********* ****, ****/*** ***** sunglasses, *** *****. *******, **** ***** that ***** ******** ******** ********** *******, <50% ** ******* **** ******** ****** accessory-based ********.
** *** ******* *****, * ****** for "***," ******* *** ****** ** people ******* ****, **** *** ******** of ******* ***** **********:
**** ***** ******* ******* *** "***," with ********* ********** ****** ******** ** "has ***:"
Direction-Based ****** ********
**** ******* ***** **** *********-***** ******** were >**% ********. ** *** ******* below, * ****** *** "**" ******* only * ********** ******:
Confidence ********** ************ *** ****** **********
****** ***** ******** *** ******** ** "Medium" ********** ******. *******, ***** *** adjust ********** ****** ** ****** **** general ** ******* ******* *** ********** search **********.
** **** *****, ********** ******* **** reduced **** ********** ****** **** ******* from "***" ** "****." ** *** example *****, *** *** *** ********** results **** ******* **** *** ********** was ******* **** "***" ** "****" in * ****** *** "******:"
*******, ******** *** ********** ****** *** not ********* *** ******. ** *** example *****, ********* *** "***" **** high ********** ***** ******** ********** *******:
***********, ******** *** ********** ***** *** not ****** *** ****** ** ******* omitted.
** *** ******* *****, "****" *** "low" ********** ********* ********* * ******* number ** *******. *******, *** ******** of ******* **** ***** ******* **** the ******-***** ****** *** ** *** requirement ** **** *********:
Reports ********* ** ****** ***** ******
***** **** ************** *** **** ******** through *** ****** ******, *** *-*** Active ***** ****** *** * ********* function **** ****** ***** ** ******** charts/track ******** ****** **** ****.
***** *** ***** * ****** ** multiple ******** ****** *** ******* **** across *******:
Server/VMS *************
*-*** ****** ***** *** **** ******* to *********/******* **** ***** ****** *******. During *******, **** ******** *-*** ******* to ******* ****** *****, ***** ******** the ******* ** ********* *** ****** from *** ****** ***** ****** *** the ******* ** ******* ************.
** *** ****** ***** **** * VMS, **** *** ****** ***** ****** and ****** ****** *** ********:
Camera *************
** *** ****** ***** **** *-*** cameras, ***** **** ******** *-*** **** to *** ****** ***** *** *-*** Configuration **** (***).
**** **** *** *********, ***** *** configure ********* *****/******* ***** *** **** app/camera ** *** ** ****:
***** ****** ***** ********** ****** ************** to ********* *** ******* ****** ** apps **** *** *** ** * single ******. ** *** ******* *****, the ****** *** *** ** ** 3 ** **** **************:
Bug ***** ** ********* ******
**** *** *** *** ****** *********** licenses **** *-*** *** **** ****. However, *-*** ****** ***** ***** **** the ****** ** *** ******** ***** to * *********:
*******, ********** ** ** ** ****** Milestone ** **** *** ********* ***** message *** **** **** * ***** throughout **** *******:
*-*** ********* **** ***** *** ** addressing **** ** * ****** ***:
** **** **** ********** **** ** an ***** *** ** **** ********* it **** * ****** ***.
Versions ****:
*******
- *-*** ******: *.**
- *-*** ******-***: *.**
- *-*** ******: *.**
***
- *********: **.*