Video analytics deployed in a public square of Mannerheim, Germany, regularly misclassify people hugging as people "strangling" each other.
According to a report from Algorithm Watch:
[******] ******** ***** ****** *** “** surveillance” ******* ***, ******* ** **** movements **** *******. ***, ***algorithms *** ******** **** ** "**********". [emphasis added]
(****,********* ******* ** ******** ************ **** ******** EU***** *** ************* **********.)
** **** ****, *** ***** ********* *** *** ******** by * ******* **** ******.******* ************* ******** *********, *** ** ************* *********, * ***-****** ***** ***** ****** "*** *****’* ******* ******* ******** ************."
*** ********** ********* ********** ****** **** video ********* *******,****** *ü****,**** ********* ***** **** ********* **** to ****** ** "*** *********** * problem" ******* **** ***** ***** ** assaults:
** **** **** **** ** *** necessarily * *******, ******* * *** can ** ***-**********: "******* * ******** hugs *** **** ******,” ** ****. “Here, ** **** *** ****** ** the ********-***** ** **** * ****.”
*******, *ü**** **** *** ******* *** clear ***** **** ******* ****** ** investigate **** ****** ********* ** ********** escalates **********. "****** ******** ** ******* generally ***** * ******", ********* ******** ***** *********** **.
***** ********* *** ********* *** ******** to ****** **** ********. *** **** on ***** ********* ********, *** ****'******* ********* ******** ****.
**** ** *** ***** ***** '******** detection' *********?
**'* * ******* **** ** *** ready *** ********** ***. **** ***'* an ******** ******** ** *** *****.
*** ***** ***** ** ************, ***** analytics **** **** ********* **** ******* to *** ***'* **** **** ***** context ******, *********** ********* *** ***** factoring ** *** ****** **** **** up ** *** ****** ******* *****.
***** ** ******** **** ********** ** "video *********", ******* **** *** *** streams ** *****, ***** *** **** many *** **** ****** ** "***** analytics", ******* **** ** *********, ** very ***** ********* ** ******.
** **** *** **** *** (********* manufactured) ******** ** *** ***** ***** to ********** ****** **** ******* ******, or *** ***** ***** *** ** a ***** ********, ***** ***** *** of ******* *** ***** ********** * very ********* *********** ** *** ****** scene ** *****. ***** ********* ******* suffers **** **** ** *** **** problem, * ****** *****, ** ***** group ** ******, *** ***** ****** to ** ********* **** ******* * general ******* **** ** ******, **** in ******* **'* **** * ****** encounter ******* *** ******.