Axis Transition To Deep Learning Video Analytics Examined
Axis launched its ARTPEC-8 chip last year with the goal of transitioning into "deep-learning analytics on the edge", but how far have they come since this announcement?
In this report, based on information provided by Axis and IPVM research, we examine the progress, timeline, and outlook on Axis's transition to onboard deep-learning analytics.
DLPU *********** ******** **** ****
**** ******* *** "******-**-**** ******* ************" into *** ******, ******* ******** ********** Unit (****) *** **** ******** ********** Unit (****). *** ** *** ********* camera ****** ********* ***** **** ******* ********:
- ** **** ****
- ** **** ****
- *** ******* ******* ******* **** ** DLPU
****'* ******* ** ****' **** ************* ************* ******** ********* *** ************** of ****** *** ******* ******** ** MLPU *******. **** ************ ** ********** power **** ***** *****-***** ********** ** develop ********** *** **-****** ********* **** Axis ********.
All *** ******** **** **** ****-********
**** ****, "** ********** **** ************ in *** ***** ******* ****** ** fixed/bullet/fixed **** *********, ***** ***** **** a *********** ***** ** *** **** units ** *** **** * ****."
*** ** *** *, *, *** Q **** ******* ** ****' ******* lineup, **** ~**% **** **** ** onboard **** ******** ********** **** (****). According ****** *************, *** ** *** * *** Q **** ******, ** ***** ** capabilities, **** ** ********* ******-* ********. Out ** ** *-**** *******, ** come **** ****, **** *** ******** with ********* ****, ****** *** *** new *** ****** ***** ******* *** ARTPEC-8 *** ********-********** ***** *** ******-*.
*******, ***** **** ~**% ** ******* include **** ***** ** *** *******, most ****** ****** ***** *** ******* the *********, **** ** ***, ***, M42, ***, *** *** *****, ** well ** *** *** *** ****** cameras, *** *** *** *******. *** majority ** ****** ***** ** *** include * **** *** **** ***********, such ** *******, ****** ****/***** ****, or ******* ******* (****** ********-*** ******** * ****).
Specialty **** ****** ******
**** ***** **** **** **** **** launch **** **** ****** ** **** specialty *****, **** ** ******* *** fisheye *******:
** **** **** ****** ******* ******* with **** ** **** ** ********** fisheye *** *** *******. ***** **** continue ** ** ******** ******* * DLPU *** **** **** (**** ******** that *** **** ** ** ********), but *** ********* ******* * ******** portfolio ** **** ******* ** *****.
Majority **** ****** *** ******-*
**** **** *** ******** ** ******** use ****** *****, **** **** ********** that *** *********, ********* ** *** M ******:
** *** ****** ** **** ** the ******** ** *** *********. *******, a ****** **** ***’* ** ********* for ***** ******** **********, ** ** also *** **** ***** **** ********* chosen ******** *********. **** ***** ** the *********** ** ******* * **** comprehensive ***** ** *********** ******** **** a **** ******* ** ********. ***** Axis ******* ********** ** ***, ** based ** *** **** ********* ******. Today, ** ********* *** ********* ***** in *** *-**** ***** ** ********, even ****** ***** *** **** *-**** products ***** ******.
Performance ********** ******* ******-* *** ****
**** ******* ****** *** *********** **** the ******-* **** ******** ** **** for **** ******** ***********:
******** **** ** *** **** ********* to *** **** ***********. ****** ** running ** ********* ************** ** ********* products **** ******** ********* ***********. *** DLPU *********** ** **** ******** ** which ****** *** ***, ** **** models *** ****** ** ********** ** specific **** ********. *** * **** rough ******** ** **** ******-* ***** between *,* *** * ***** *** performance ** **** *** ********* ********.
Avigilon ***** ** **** ** ********* *******
***** *** ****** ********** ** **** learning ******** ****' *********** *** *********, Avigilon ******* ***** ** ***** ** product ******* **** **** ******** *** performance. ****** ****' **** ******** ******** to ***** ************ ********, ******** ****** a *******-********** ******* *** *** ** its ***-**-*** ********. **** ****** ******** products **** **** * ***** *******, it ****** **** *********** *** *********** across *** **** ******** ******* *** analytics.
**** ******* ********** *** ************** ****** ******** ********* ************** ** people *** ********, *** *** ** appliance ****** * *********** ******* ************* ******, ******* **************, *** **** recognition.******** ******** * ******* ** ******** analytics ******************* ******,****** ***********,******* ******** *********, ********* ********** **************** *** ***.
** **** *********** ** **** ******** within *** ********, **** ******* **** to ******** *** ****** ***** *** cameras **** ******* *********, ********* **** Avigilon. **** ****** *********** ********** **** as*** *** *********** ** *********** *******, **** *** ********* **** a *** ***** *** ********* *******.
Hanwha *** **** ******* *******
****** ****** ********* ******* *** ******* products, *** ******* ** ****, ****** is ************ ****** *** **-****** *********. Hanwha ** **** ************* ** **** learning, **** *** ********, *** **** a ***** ******* ** *** ******* lineup *** **** ******** *******. **** Axis *** ****** **** ** **** strides ****** **** ******** ********** ** compete **** **** ******* ***** ** companies **** ****** ********* **** ********.
~**% ** *********** ******** ****** ** *** *** *********** to ****, ********* ******** ****** *********** between ***** *** ********* *** *** transition ** **** ********.
** ******, ****** *** ******** **** out. ** *** ***** ** *** M32s (** ******** **** ******* ****), but *** *** *** **** ** the**** ******* ************* **** **** ** ** ***** to **** ** ******** **** **** last ****. **** *** ****** *** and ********* **** ******-*:
*'** ******* *** ******** *******:
*** ** ** *-**** *******, ** come **** ****, **** *** ******** with ********* ****, ****** *** *** new *** ****** ***** ******* *** ARTPEC-8 *** ********-********** ***** *** ******-*.
** ******,
****** *** *** ***********. ** ***** stated, **** ******** ** ***** ******* ********* ****, *** ** *** **** **** were *** ******. **** *****'* ****** to *** ****, ** ****** ** current.
*** ********* **** *** *** *** DLPU-equipped *-****** ** ***** ** **** is *** **** *******. ***** *** M3216, ******** ***-********, **** *** ** ARTPEC-8 (***** ******** * ****). **** should ** ************ ********** ** *** current ***** *** *****, ****** ***-***** (and *** ******* ** *** *** the ******-*** ******).