Uniview Deep Learning Camera Tested

By Ethan Ace and Rob Kilpatrick, Published Jul 14, 2020, 01:33pm EDT

Uniview's intrusion analytics have performed poorly in our shootouts. Now, they have released their Prime Series Deep Learning cameras, with an "enhanced detection algorithm", but how well do these new models work?

IPVM Image

We bought and tested the IPC264SA-DZK, examining:

  • How accurate are person and vehicle classification?
  • Does it detect people when crawling, running, or obstructed?
  • Does it false alarm on common outdoor sources?
  • Does it ignore common indoor sources of false alarms?
  • How far can it detect people?
  • Does it integrate to third-party VMSes?

*******

***** *** ********-*** ****** missed ****** ** ******** in *** ***** *** did *** ******* ** small *******, ****, ** blowing *******, ** ******** from ***** ***** ****** similar ** ***** ** models ** **** ****** (such *******************), **** ***** ******* and *** ****** ************ detected *** ********** ** human, *** ********** ******** re-triggering ****** ********** *******, despite *** ******.

Vs. ***** *********+ / *** *** ********* ********

*******'* *********** ** ***** compared ** ********* ****** competitive ******** ****** **** Dahua *********+, ***, ** Hikvision ********.

  • *******'* **** ******** ****** performed ********* *******'* *********+, ***** **** ************ classified ***** ******* ** people *** *********** ** parked ********. *******, *******'* weak *** *********** *** classification ** ******* ***** as ****** **** ****** not ******* ** *********+.
  • *******'* ** ********* ****** than*****'* ***** ****** ********* (SMD), ***** ******* ** small/large ******* *** ******, with **** ***** ********* range (~** *** **. 11 ** *******).
  • *********'* ******** ******* *** NVR******** **** **** ***** alerts ***** *** *** trigger *** ******* ******, including ***** ******* *** shadows/reflections. ******* ** *******, AcuSense ********** *********** ** parked ********, *** *** not ****** ******* ***** as ******.

No ****** ******* / ****** *** ******** **********

** *** *******, ****** and ******** **** ********** detected, **** ** ****** objects (****** **** ************** issues, ********* *****). ***** subjects **** ********** ******** when *******, ********, *** running, ***** ****:

IPVM Image

Wheels ********** ** ******

****** *******, *** ******* AI ****** ********* ********* on *** ******, *********** them ** ******, ***** below:

**** ******** *** *** seen ** *** ***** person ********* ******** ** have ****** (****** ** issue ****** **** *** *******).

Large ******* ********** ** ******

*******, ***** ******* (**** as *** ***, *****) were ************ ********** ** people ** *** *****, triggering ***** ******.

*** ******** *** *** trigger ** ***** ******* in *** ***** ** view, **** ** ****, squirrels, *******, ** *****.

No ****** ** ****

*******'* ** *** *** trigger ** ***** ****, throughout ***** ******** ****** in *** *******.

No ****** ** ******* / *******

** *** *****, *** Uniview ** ****** *** not ***** ** *******/****** foliage ** *******, *** most ****** ******* ****** of ***** ******.

No ****** ** ******* / ****** *******

*** ******* ** ****** did *** ***** ** common ******* ** ****** false ******, ******. *** example, ****** ******* **/*** and ****** ******* *** not ******* ***** ******, common ****** ** **** other ** ******* *** non-AI *********.

Person ********* ** ~** ***

** *** *****, *** Uniview ** ****** ******** humans ** ~** ***/~**' distance ***** * ~**-**° AoV. **** *** *********** is ******* ** *********** such ** **** ****** (*** *** ********) ********** **,***** ***/***,***** *********+, ************ ********/**********

IPVM Image

No ***** ***** ******** ***** ***********

** *** *******, *** IPC264SA-DZK's ******** ****** *** not ********* **** ***** party *****, ********* ********, Exacq, *******, *********, *** Network *****. ** ****** nor ******** ***** **** sent ** ***** *****.

Similar **** / ************ ** ******* *******

*** ********-*** ** ******* size *** ************ ** competitive ****** *******, **** as ***** *** *********, shown ****, **** *** using ***** ***** *** similar ******:

IPVM Image

*******

*** ********-*** *** ** estimated ****** ***** ** ~$300-325 ***, ******* ** Dahua's *** (*******,~$*** ***) ********* ********(**-*********-***,~$*** ***).

Uniview **** ******** ****** *** *********

*** ******* ********-*** ****** uses * ********* ***, seen *****.

IPVM Image

Versions ****

*** ********* ******** ******* was **** ** *** tests, *** ****** ** the **** ** *******:

  • ******* ********-***:*********-*************

******* ******* ** **** a *** ******** ******* is ********* ** ***********, but *** ********* ** the **** ** *******, which **** ******* **** improve ************** ***********. ** will ****** **** **** new ******** **** *********.

Comments (11)

Undisclosed #1

IPVMU Certified | 07/14/20 01:54pm

******* ~$***, ********* ~$***, Dahua ~$***

IPVM Image

******* ********** ** *** best.

Undisclosed #2

In reply to Undisclosed #1 | 07/14/20 06:39pm

**** ** * ***** comment **** ****** ** purpose, *** **** * cynical ****.

*** ****** ** **** point ***** **** ** what *** ***** ** willing ** ***, *** many ******* **** **** way, *** **** *** Chinese.

Michael Miller

In reply to Undisclosed #2 | 07/14/20 08:20pm

**** **** **** *** because ***** * ********* OEM *** **** ***+ companies.

Undisclosed #1

IPVMU Certified | In reply to Undisclosed #2 | 07/16/20 04:45pm

*** ****** ** **** point ***** **** ** what *** ***** ** willing ** ***, *** many ******* **** **** way, *** **** *** Chinese.

**** ****** **** * rational ***********: ****** ****** forces, *.*. *****/*************, ***** similar ******* ********.

****** * *** **** C.E.:

********** ********* ** *** process ** ***** ********* or ********* ******* ********* evolve ******* **** *****, coloration, ******, *** ***********.

IPVM Image

Avatar

Skip Cusack

07/15/20 07:10am

*****,

*********** *******, ** ******. Presumably **** ******'* ********* performance ****** ******* **** time. **** ** ******* itself? **** *** ************ set ************ ** **** of ***********, **** ** steady *****, ***.? **** AI ******** ***** * lot ** ***** *** is ******** *** *** all ****** ********* **** go ** ****.

Avatar

Ethan Ace

IPVM | In reply to Skip Cusack | 07/15/20 10:19am

**, ** **** *** retrain ******. ********* ***** actually ****-***** *** ****** rare.********'* **/** ****-******** ************ ******** *** **** well-known *******.

*** ******** ** **** "deep ********" ******* ***** from * ***-******* ****** network, ***** ** **** in ***********, *** ******* in *** ******(*) ******* firmware *******.

* ***** ***** **** you **** ** ** likely ** ******* "********" analytics **** ****. **'* getting **** *** **** common, *** ***** *** a ****** ** ******* with ****** *********** ** this ***** *****. ** it ** *** ***** of ***-********** *********? *** yet, *** ******** ** standard *** *********** *********, * **** ***********.

faheem adam

08/05/20 03:49pm

***** ******* ** ** many ******* *** **, analytics *** *** ************ of ***** **********. **** the ************* **** ******* out ** *** **** (IPVM) ** *** ****** and **** *** *** to ****** **** ******* updated ******** ** ******* algorithms ** ***** *** other *******? ** ** they **** *** **** there ** ** **** false *********.

John Honovich

IPVM | In reply to faheem adam | 08/05/20 04:00pm

******, ******* ************* **** us ******** *** ******* firmware. **'* ****** ****** a **** ****** ** IPVM **** **** **** we **** ************* *** preliminary ******* **** ********, they ******* **** **** variant ** (* ** only ******** ***** *********) "Oh, **, **, **'** get *** **** ******** now."

faheem adam

In reply to John Honovich | 08/06/20 01:48am

***, ** **** ***** what ******* **** *** give ********* ***********, ****** and ******* ******** *** free ** ****. *** need ** ** ** them *** ****** ** with * **** ***** tag **** **** ***** have ***** *** *******.

* **** *** ***** same ***** **** *******. no ****** *** **** examples *** **** * do ** ***** ********* ptz ** **** **** the ************** ** *** algorithm *** *** ** get **** ** *** it, *** **** *** my ******* *** *** theirs ** ****. ** go ** ** *****.

******* **** ** *** good ****

Undisclosed #3

08/19/20 08:20pm

**** *** *** ** use? *********?

Avatar

Rob Kilpatrick

IPVM | IPVMU Certified | In reply to Undisclosed #3 | 08/24/20 12:41pm

***, *** ********-*** **** a ********* ***, * added * ******* ** the ****** **** *** SoC ****.

Read this IPVM report for free.

This article is part of IPVM's 6,728 reports, 907 tests and is only available to members. To get a one-time preview of our work, enter your work email to access the full article.

Already a member? Login here | Join now
Loading Related Reports