Watrix Gait Recognition Profile

By Zach Segal, Published Oct 16, 2020, 10:36am EDT

Watrix is the world's only gait recognition surveillance provider IPVM has found. And, inside of China, Watrix has gained significant attention for its claimed capabilities.

IPVM Image

Gait recognition interest has increased as coronavirus and privacy concerns have raised concerns about facial recognition usage.

For background, see IPVM's Gait Recognition Examined and Single Frame Gait Recognition From Michigan State and Osaka University Examined.

Inside this note, we examine Watrix and what they are doing inside China.

Executive *******

******** * ******* ***** computer ****** ******* **** claims ***** **** *********** is ***** ********, **%, working ****** * **** range ** ****** *** at * ******* ******** than **** ***********. ****** says **** ******* **** recognition ** ******* ****** across *****.

*******, **** ** ********* as *** ******* *** not ******** **** ******** evidence ***** *** ******* or ****-***** ***********. ****, our ****** ** *** academic ********** ***** ***** on **** ** ***** claims.

****** ******** ** ***** with ****.

Watrix ********

******** * ******* ******** vision ******* ************* ** Beijing **** ********* ************************ ** *** ****** in *** ******'* ********** *****. ** *** founded ** **** ****. ******** ******** ********* ** ************ ** **********, ******* Academy ** ********(*****) ***** ********* ******** **** *********. ** ****,****** ******* ** **** raised $**.* ********** **** *** ***** no ********** ** *********** since ****. *** *******,*** *** ********** *** been **************** ** ************ ******* and ********* ** **** of ***** ***** **** initiative ** ***** *** of ***** **** *******.

**** ***** **** ** Watrix ********* ** *********** outside ** *** ***. If ***** *** ***, we ***** ****** ** to ** *******.

**** ***** **** ****** captioning **** ******* ******* on ******'* **** ***********

Production ******* **** **** / ************

*** ****** ** **** Watrix ****** ****** ************* of *** ***** ******* actually ***** *** *** one **** ** *** found, ****** *****, ****** questions.

IPVM Image

*** ******* *** **** the ***** ********* *** 2020, ** *** ****, and **** ** *** right, *** *** ****** of *** ****** ******* to ** ******* *** the ****, ******* ** being * ***** *****, as ***** *****:

IPVM Image

All ** *** ** ******

IPVM Image

****** *** ********* ******* the "**** ** **** ****" ***** ** ** all-in-one (***) ** *** investigation. ****** **** ***** upload ******* ** *** device *** **** **** ~10 ******* *** ***** hour ** *******:

***** **** ** ****** videos **** *** ******* first *** ***** *** machine ** ******* ***** of *** ******, **** the ******* *** ******** a ****** ******* *** hour ** ***** ** 10 *******.

**** *********** ** **** more *************** ********* **** other ********** ******* ** requires ******** ****** (~* second ** *******) ******* of * ****** *****, increasing *******.

********** **** ********* ******** *******, **** Dahua *** *********:

*** ********* ******* *** also *** **** ******ordinary ******* on the market, such as ordinary cameras developed by manufacturers such as Hikvision and Dahua

******** ****** ******** *** reduce **** ********, ***** significantly. ******* ***** *** set-up ** ******* **** images, ******* **** ***** be ***** ** ******* not ************* ** **. Cameras *** **** ***** placed **** **, ***** can ***** ***** ****** that ****** **** *********** accuracy. ** ********, **** recognition ******* ******** ******** gait ******, **** *** right *********, *** **** can ** **** ** capture ** **** ******* areas ***** ****** *** often ********* ** ***** people, ********, ***.

Watrix ********** **** + **** *********** ****** *********

IPVM Image

****** ******** ************* **** *** ***** face *** **** ******** for ***********. **** *********** performs **** ** ***** frontal ****** ** *** face ***** **** *********** is **** ** ******* images. **** ***** ***** one ****** *** **** challenging, ** *** ******* captured ***** **** ** good *** *** **** of ***********. ****, ********* different ********** ********* *** computation *** ********** ****** of ********* *** ***** cause ************. **********,***** ********* ** ***************** *** ****** ******** and ******.

Watrix ****** **** **% **** ***** ********

****** ******** ************* *** ******* **% accuracy. ******'* ***** ** 94% ********, ** ****** than ****'********* ************* ** **** recognition*** **-****************** **** ******* ******************* ***** **** *********** algorithms *** ******* ** the ****-*****. **** ******* with *****, ******** ********* (ie *****, *******, **** worked ***), **** *********, and ******** *******. ** addition, **** *********** ** adversely ******** ** ******** changes *** ****** ******. IPVM ** ********* **** they *** ******* **% recognition **** ********, ****, road, ******* *******, *** other **********.

Advertises *********** ** * ********

****** ********** ************* ** ** *** but *** *** ******* a *** ** ***. Gait *********** **** **** ******** ******** ***** *** than **** *********** *** can ** **** **** 15-20 ***.

Watrix ********* ******* ** ******

****** *********** ********* * variety ** ******.**** *********** ******** ** greatly ******** ** ********** angles **** ******* *** face ** **** ******* of * **** **** and ** ******** ** harsh ******** ****** *** face *********** **** *** work ** * *******'* back. **** ****. ***** ********'******* ************ ***** **% accuracy **** ***** ******.

IPVM Image

Watrix ****** **** ********** **********

****** ******** **** ***** gait ********************* ******* ******** * specific ***-** *** ***********. Watrix ***** **** ****** cameras *** ***-** ******** for **********, *** ******** then **** ****** *** designated **** *** ** least *** ***** ***** an ******** ******* *** footage, ** ***** ***** below *****:

IPVM Image

**** ********** ** ****** than ****** *********** ***** watchlist ****** ***** *** be ***** **** ** databases ** ****** *****.

****** *********, ** *****, implies **** **** *** capture **********'* **** **** surveillance *****, ******* ***********, but **** *********** ***** instructions ***** ***** ********** an **********'* ****. ** believe ** ** ******** to ******* ** **********'* gait *************** *** *** impact ** ********* ** matching ******** ** **** to ** * ****** in **** * *******.

Technique *** ******** *********

****** *** *** ********* its **** *********** ****** and ****** *** ********* ******** * ******* ** techniques. ****** *** *** said *** **** ** gait *********** *** **** probably **** ******* ***** methods ***** ***** ******** in ****.**. ***** *********** ********* ****** **************/***-*****,*****-*****, ******-**-*** ***************, ** *** ** those ********** ** ****** could ** **** ** Watrix. ****** **** **** not ***** *** **** and **** *** ********.

Claim *****-* ******

****** **** **** **** set * *********** ** *** *****-* gait *******. ****** **** they **** *****'* *****-* results **** **% ******** in * “******” *******, 96% ** ******** **** backpacks, *** **% ** those **** *****.

IPVM Image

IPVM Image

**** ******* ** ************* easier **** ****-***** **********.********-******** ** ***** ** ~head ****** ** ****** subjects **** * ***** background **** ******** ******, with **** *** ****** (coat ** ********) ****** at * ****. **** in **** ***** *********, with ******* ******* * coat, ******'* ******** ******** dropped **** **% ** 91%.

** ** ******* **** Watrix's *********** ** **** world ********** **** ** lower **** **** ****. What ****** ** ********** is *** **** ***** it ***** **, *.*., when ******** * ******** with * ****** ********** at ********** ****** *** tracking * ******* **** a ******, ********, ***.

Used *** ************* *** ************

****** **** ********** ** being **** ** * few ****** *** ************* and/or ****-**** ************ ********* to ******** *** ******.

  • ** ********,**** *** ******* ************ ** ******** * gait ******** ** ******** people
  • ** *****,*** **** ******* ********** ** ******** * gait + **** ******** of ******** *** ******** people **** ***** ** used *** ****-**** ************
  • ***********, *** ***************** ****** **** *********** is ***** **** ** investigate *** ****** ***** there ** ** ****** information

** ********,****** **** *** ************ ***** **** ** a ************ *** **** for ****-**** *** ************* by *** ******* ****** and **********. *** **** it *** **** **** in ** *****, ********** over * ** ** public *******.** ******* ****** **** in **** ******** ** ************ ** sell ~$** ******* ** technology ** ********* ****** China.**** ***** ******** ******:

** *** ********* ******* with ****** ** ******** investigations, **** ** ******** suspects **** * ******* scene, ... *********, ***** has ***** ***,*** ****** criminals ** *** ***** and ********. [*** ********’*] database ******** ***** **** a ***** **** ******.

IPVM Image

** ***** *** **** any ****** ********* ********** the ******** **** ************ being *********.

Expansion ** ***-******** *****

****** *** **** ******** started *********** ********** *********, including*********** *********”, * ******* ** inspect **** *** *******.

IPVM Image

**** ** ************* ******* from ***** **** **** / ****** ****** ************. Often **** ********* *** struggling ** ***** **** market, **** **** **** expansions **** ***** ********* ones ** ******** *******. We ****** ** ******* of ******'* ******** ****** in **** *********.

Watrix *******

***** ** ****** ** certain *** **** ****** performs, ** *** *********, because ** ***** *** lack ** ******** ********, the *********** ********** ** conducting ****-***** **** *********** and *** **** ** competitors, ** ** ******* of *****, ******** ****.

*********, ** **** *********** can ** **** ** work ******** ** *** real *****, **** ******** organizations ***** *** **, but *** ****** ** Watrix ** **** *********** remains ** ** ****.

Comments (1)

****** *** *** **-***** analysis.

********* *** **** ** demo ********* *** ******: Back ** ***** **** the ***** ***** ******* Post *********** ******* ***** *********** **** ******** * short ***** ***** ***** contains **** *********** ******** from **** ******* ** be * *** ***** and ***** **** *** demo *********:

*** ***** ****'* *********** is **** *** ****-**** seems ** *** ** what ******* ** ** a *****-***** ******, ******** Ubuntu ** * ******* distribution ***** *** ***** desktop ******* (*********):

IPVM Image

** **** **** *** can **** *** *** background/foreground ************ ********** ***** leads ** *** ********* silhouette ** *** ******:

IPVM Image

*** ****** ** *** segment ***** *** *****-********** of *******-******* ****'* **** mentioned ** **** ******* with *** ************ ***** shown ** *** ***** and *** ********* ****** shown ** * **** at *** ******. ****'* slightly *** **** ** that **** *** ********** extraction *** **** ** the ****** ******** *** these ********** ******* *** cyclists *** * ******** doubt **** ***** ***** can **** ** ********:

IPVM Image

* ****** ***** ** get ** *** **** of *** *********** ********* with * ********* *** filename ***** ** *** top (********** *** **** that's ***** ********) *** some **** ** ****** below **** ******** ****** (maybe **** **** ** confidence ***** ** ******** by **** ******* ******** algorithms):

IPVM Image

**** ******** ********** ***** is ***** **** ******* directly **** ** * bounding *** ** * pedestrian ***** ** **** labeled ** "******" (** is *** ***** ** the **** ********* ** the *****):

IPVM Image

******* *********** ****** **** is *** ****** *********** in *** ****** ***** shows *** ********** ***** which ****** ******* ***** and *****. ** **** no **** ** **** fps **** ***** *** recorded *** ***** **** according ** **** ******** the *** ************* ** needs ***** *** ***** of ******* (** * 6x ******* **. ****-****) the ***** ** ***** processing ***** ******* ** as ****** ***.

IPVM Image

************* *** *** ** later ****** ** ****** that ***** ******* ******* "can *******live ****** **** ** ** ***,*** *******".

IPVM Image

* ****** ***** ** get ** *** * pedestrian **** * ******** box ****** **'* ******* whether **** ***** **** a ******** ****** *** recognized ** ********* **** entirely. (** ******* ******* from *** **** ****** there *** **** ***** pedestrians *** *** ********** but ***** ******* * bounding ***.)

IPVM Image

**** **** ** *** mean?

* ** ** ** looks **** *** *******'* focus ** *** *** mostly ** *******-***** ********** but **** **** *** at ***** ****** ** move ******* ** **** already **** **** ********** as ****.

* **** **** ** can **** ** *** video ***** *** ****** front-end ** *** *********** does **** ****** ************* and *** **** ** early *********.

* ** ****** **** no **** **** *** accuracy ** **** ***** is *** (*) ***** trust *** ********* *** (b) *** ********* ** all ***** ********** ******* and ******** **** ****** me ** ******** *** and * *** **** finely ***** ********.

Read this IPVM report for free.

This article is part of IPVM's 6,651 reports, 895 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