I recently installed 3 of the new Axis deep-learning cameras (1 P3255 and 2 P3265) cameras using Object Analytics, and overall the performance has been excellent and the results match those of the IPVM testing.
However, one significant surprise has been that all of these (almost universally) miss a straight-on approaching vehicle at night (IR illumination being used). The problem is significant enough that we're considering deploying beam sensors or similar to provide an alert for this type of scenario.
****'* ** ******* ** * ******* capture **** ***** *** ******** ****:
*** ****'* ** ******* ** * night-time ****** *******:
** ****: *** ********** ** *** because **'* *** ****** ***** **** triggers ****-********** *********. **** ** *** same ****** *** **** ******** ** the *** *****. ** **** **********, the ***** ******* ** **** ** night.
** **** ********* **** *** ****** / ***** ** ******? **** ****** have **** *********** ** *** ** improve ****? **'** ***** ********** ***** time ******** ** *** *** *** a ****** ***** *** ********* - but ** **** ** ***. *** ideas ***** ** *****. ******.
* **** *** **** **** ****** though **** ** ** ********** **** object ********* *** ** **** *** industrial ***** ***** *** ** ***** kicked **. **** *** ******* *** to **** ** ** *** **** contacts **** ****? * **** **** sometimes ***** *** ** ***** ******** so *** ***** **** ****** *** are ** *** ****** ******* ** firmware.