Dahua Emotion Analytics Tested

By Rob Kilpatrick, Published Dec 21, 2020, 11:02am EST

The advance of AI promises more sophisticated recognition. For example, Dahua markets the ability to distinguish between angry, calm, confused, disgusted, happy, sad, scared, and surprised.

IPVM Image

We tested the Dahua IPC-HDBW7442H-IPZ running these analytics, examining:

  • How accurate is emotion detection?
  • Does accuracy vary depending on emotion?
  • How does the angle of incident impact accuracy?
  • How does distance/PPF impact accuracy?
  • Does facial hair impact accuracy?
  • Do face coverings affect detection?

*******

*****'* ******* ********* **** frequent ********, ****** *** most ****** ************** *** 'calm' *** ****** *** had *******/************** *****, ***** is *** **** ****** state *** ******. ***** and ******** **** ******** most ** *** ****, while *****, *********, ***, scared, *** ********* ****** poorly.

**** ***** ******** ** overview ** *** *********** for **** ******** ***** offered:

IPVM Image

******* *** ****** ******** to ******** ******** *** time * **** ** detected, **** **** *******, facing ****, ** *** PPF, ***** **** *** tendency ** ******** ******** as "****" ** ***** instances, *** ******** *** appear ********, ***** **** subjects *** ****** ** be ************** **** ** the ****.

Best ******** ** **** ***/******* ***** (*** ***** *********** ******)

** *** *******, ** began **** * ****** scene, **** ***, ** a *** ***** ** incidence, **** ******** ****** the ****** ** ***** range ** ********* ******** factors **** ** ***** angles ** *** ***.

**** ** ***** **********, emotion ********* *** ****** inaccurate. *** **** *********** consistently ********** ******** **** Happy:

IPVM Image

*** **** (******** *** camera ********** ***** ** Calm ***** **** **********, discussed *****):

IPVM Image

*******, ***** *********** **** as *****, ***, *********, or ****** **** **** often ******** ** ****, or ********* *****.

IPVM Image

Accuracy ***** ** ******* ****** *** ***** ***

** ****** ** ********* greater **** ~**°, **** subjects **** ********** ** Calm, ********** ** ***** actual ****** **********.

IPVM Image

*** ****** **** ********** the ***** ** ********' heads ** **** ****** testing:

IPVM Image

*********, ** *** ****, subjects **** **** ********** as **** ** *****.

IPVM Image

Incorrect ************** **** **** ********

***** *** ****** ******** to ******** ******** *** time * **** ** detected, ** ********* ** classify **** **** **** of *** *******'* **** obscured. *** *******, **** partially ******* ** *********, the ****** ***** ********** the *******, *** **** as **** *** *****.

IPVM Image

Reduced ******** ** ****** **** ****** ****

****** *******, ****** **** facial **** ********** * facial ******* **** ****** and ******* ****** ********** calm, ***** *** ****** with * ***** *** angry **** **** *** classified ** *****.

IPVM Image

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

** *** *****, ** face *** **** ********** as *** ** *****, instead, **** **** ****** classified ** ********, *****, or ****. *****, *** person *** ** ***** face *** ** ********** as ****.

IPVM Image

Confused ******** ** ****** ******* ****** ****

****** ******* ******* *** scene **** **** ******* a ***** ******** **** that *** *** ******** by ****** **** **** accurately ********.

IPVM Image

****** **** ****** **** confused *** *** ****** hair, ***** ******* ********* did *** ********** ****** the *******.

IPVM Image

Disgusted ************ *********

*** ********* ******* *** sometimes ******** ** * person ******* ******* *** scene **** * ***** disgusted **********.

IPVM Image

*******, **** ** *** time *** ********* ********** was *** ********, **** on ** *********** ******* to *** ** ** classify.

IPVM Image

Simple *****

******* ********* ***** ** simple, *** **** **** turn ** ***** ******** in ********, *** **** it **** **** *** metadata ********* *** ******* classification ** *** **** view ******.

Tech ******* *** ***** ********

***** *** **** ******* tried *** *** ****** to **** ******* ******* detection, ****** ******* * clearer **** ** *** face, ***** ***** ********** did *** ***** ** our ******* ** ******* on *** ***********.

*******

*** ***** ***-*********-*** *** be ***** ****** *** ~$920 ***.

Versions ****

*** ********* ******** **** used ****** *******.

  • ***** ***-*********-***: **.***.*******.*.*, ***** Date: ****-**-**

Comments (11)

Undisclosed Manufacturer #1

12/21/20 06:17pm

*** *** **** ** me, * ***'* ***** of * **** ***** use **** ** *** video ******** ***** ***** these ********* ***** ******* a **** ********** ** investigative/forensic *********. ************ ** light ** *** ************ and ************* ** ****** someone's **** ********* ***** from * ***** ******. I ** "*********" *** "disgusted" ** *** **** time **** ********** **** I'd ******. *** **** that *** ******** ***************?

***** *** ** *** cases ** ************** (*** detection *******) ** ********* analysis *** ****'* ***** it...I *****. *** ****'* based ** **** ******** as **** ** ** is ****** **********.

*** **** *'* *******.

Andrew Myers

In reply to Undisclosed Manufacturer #1 | 12/21/20 08:04pm

****'* * ****** ****:

** ***'** * ***** invasive ********, *** ***** use **** ** ***** profiles ** ******** *** more ********** ******* **** they ****.

** **** **** ******** to****************, *** ***** ** able ** ******* ** someone ** * ***** was ***** ** ***** trouble. *'* ****** **** there *** * ***** several ***** *** ***** using ********* ********* ** spot ********** ****** **** acted. (****** ***** **** of *** ******* *'** heard, *** ********** ******** could ** ******* ********* than **** ***'* ******.)

** ******, ** *** analytic ***'* ********, **** it's **** ** *** it ***** ******.

Mihai Simon

12/22/20 05:38am

**** *** *** ******** article ** * **** time... * **** *** faces (*** *** ************* misclassifications). * *** *** guys *** * ***** time *******... ***** *** for ****** ** ***** at *** *** ** 2020!

Humberto Macedo

12/22/20 07:22am

*** ***** *** **** expressions ** ******** **** they *** *** ****** feeling *** ****** *** train ** ** ** and ** ***** *** system. ** ** **** easier ** ***** ******* to ******* ** ** actor **** ** ** to ***** ******* ** fly ** ********, ******** 9/11. *** ****** ** an ****** ********* ****** that ******* **** *** deliver *** ******** *******?******* misleading. ** ** *** SECURITY. ** ** * FAIRY ****. **** ****** exists *** ***** *** tiny ********* ****** ** facial ******* **** ** NOT ****** ** *** person's ****. ***** ***** that **** ** ******** emotional *********, *** ****** INVOLUNTARILY ******** * **** muscular ******** ** *** face. **** ******** ** detected ** ****** *** are ******* ** *** subject, ** ** ******* who *** ******* **** to ******** ******'* ********* at ******** ******** *** conduct *** ****** ********* to *** *********** ** not ** *** ******, expressed ** ***** ************* and ***** *********** ****** movements. **** **** *****, it ** ****. *** ONLY ****** *** ******** these ****** *********, ********** DOES *** ** **** yet *** * **** simple ******, ******* ***** are **** **** ***,*** known ***** *********, **** one ********** * ******, which ***** **** ** be ******* ** *** algorithm ** ***** ** individualize ***** ******** *** each ******. **** ******. Easy *** ***** ** do, ***'* **?🤔

Kris Tibbitts

In reply to Humberto Macedo | 12/24/20 06:44am

***, ** ***** ** impossible ** *** ******** near ***% ********. *** even ** *** *** get * ****** **% accuracy **** **** ********** or ******* *************, ** could ******** ***** *******. If ** ******** *** violent ***, * ***** it ***** *** ** worth **.

Kris Tibbitts

12/24/20 06:37am

** ** ****, ** would ** ****** ** test **** ** **** life. ****** ** **** expressions *** ******** *** are *** ****** ****** is *** ** ******** test. * ***** **** to *** **** **** for * ****** ** time ** * **** life ********* *** **** compare **** ** ***** by ******* *** ** trained ** **** ***********. I ** ********** **** this ***** ** *********. And * ** **** it ***** *** ** much ******, *** ** would ** *** ***** way ** **** **.

John Honovich

IPVM | In reply to Kris Tibbitts | 12/24/20 01:44pm

****** ** **** *********** for ******** *** *** not ****** ****** ** not ** ******** ****

****, ** * ********** you *********, *** ******* Dahua ******** **** *** (below) *** **** ****** being ***** *** **** Dahua ******** **** *** was ******** ***** **** inside?

IPVM Image

***** * ** ********* joking, ** ***** ** that *** ** ***** video-based ******* ********* ******* are ********** ******** ** recognize *** ******* ********** of ********, *** *** actual ********* ***** ** the ******. ****'* ****, no?

** ** *****, ******* of ****, ***** ** an ******** **** ** any *****-***** ****** ***** it ******* *** ******* appearance ** '*********' ** equal ** ****** ********* when *** **** ******* people ***** **** *** appearance ** ***** ***** when **** *** *** (or **** *****).

**** ****, ** * system *********** ***'* ********** above ** *********, ** indicates ** *** **** fundamental ********.

***, ** ****** **** real-life ** **** **** various ******** ****** *** the ******** **** *** same.

Kris Tibbitts

In reply to John Honovich | 12/27/20 06:50pm

****, * ** *** saying **** ** ***** turn *** *** *********. Was *** *****? **** was *** **** **** you *** ** ***** face ******** **** **** that? ***** *** ****** was *******. (**** ****) That ** ** *****. It ** **** **** to ***** *** ****** when *** ***** ** the *** ** *** intended. * ***** ****** to *** *** **** world ******* ****** **** posed ****** ***********. ****, I ** **** ** will *** ****** **** time. *****, ** ***** near *******, *** ******. Enough ** ******** * first **** ** *******.

John Honovich

IPVM | In reply to Kris Tibbitts | 12/27/20 07:04pm

** ** **** **** to ***** *** ****** when *** ***** ** the *** ** *** intended.

****, ** ***** ** that ** *** *** it ** ********. ***** is *** ******** **** they *** **** *****. Dahua ** ******** **** when ********* **** ** expression ******************* ***** ****'****', ** '*****' ** 'surprise', ***., **** **** report *************.

IPVM Image

** **** ****, *** happy **** ********** *** lips ********** ****** ******, which ****** ********* ********* with *********. ***'* **** are ****** ********, ***** typically ** ********** **** being *******:

IPVM Image

***, *****, ** ****** it ** '**** **** scenarios' ** ****,**** *** **** *********** results, *** ** *** showing ***** ******, ***** ones ******* ** ** show * ****-**** *** where *** ** ***** but ** ** ******* away ** ** ** angle ** *** **** is ********* ********, ****** will ******** **** **'* just ******* ** *** bad ****** ***********.

****** ** ******** * first **** ** *******.

****, ** *** **** to *** **, ****'* fine. **********, **** ********* ****** ***** ********* we ******, ** ******* ** what ******** *** ***** and *** **** **** can ******** ********.

Kris Tibbitts

In reply to John Honovich | 12/27/20 02:20pm

*** ****** **** ** works... *** ****, ** emoji ** *** *********** an ****** ************** ** what *** ****** *** be ******* ***. **** you *** ******** ** the "*************" ********** *** not ** **** *** system ** ******* ***. He ***** **** ** has *** **** ** his ******** *** ** trying ** ******* ** without ******** ** ** me. * ***** **** would ** **** ** an ******* ******* ******. Regardless ** *** *** guys *** ** *** posted **** ******** *** this *******, * ***** still ****** ** ********** to **** **** ** would ** **** * period ** **** ** the **** *****. ** is ********* ********* **** you *** *** ******** so ** ** *** fair ** *** ****. This ** *** * feature **** * ***** ever **** ** *** or ****, **** *********** to *** **** ** being ***** **** **********.

* ****** ** ** saw *** ***** ** a *****. (*******) *** should **** *** ***** then *** ** *****. (joke)

John Honovich

IPVM | In reply to Kris Tibbitts | 12/27/20 02:43pm

* ****** ** ** saw *** ***** ** a *****. (*******) *** should **** *** ***** then *** ** *****. (joke)

** ********* **** ** the ******, ****** ***** for **** ***********:

****** *******, ****** **** facial **** ********** * facial ******* **** ****** and ******* ****** ********** calm

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