Ambarella CV72S Chip And Generative AI Onboard Camera Capabilities Analyzed
********** ** **** ************* ****** ***** analytics *** ********* ******* *********** ***** processing. *** **** *********'* ***** ****** the ******* ** ********** ** *** video ************?
** **** ******, ***** ** ** interview **** ********* *** **** ******* of ********** **, ** ******* *********’* latest *******, ********* *** ************.
*******, ** ******** *********'* **** *********'* **-******* ****** ******.
Executive *******
***** *** ********* **** ******* ********** AI ** *** ******* *******, *** company ************ *** ************ *** ********** will ** ******* ******** ** **** cloud *******.
*********'* ***** **** ** * *********** advancement ** ***** ********* ** ******** generative ** ********** ** *** ****, a ********** ********* ******** ** *****-***** systems. ********** *** *** ****** *.* architecture, *** ***** ****** * ** increase ** ** *********** **** *** predecessor, *** *****, ******** ****-**** *** tasks **** ******'* **** ******** ** surveillance *******, ********* ** *********. **** enhancement ******** **** ******* *** *********** in ***** ********, ********* ***** **** as ******** ********, *** ***** ******* compared ** **** ***** ***************.
************, *** **** ********** ** ***** signal ********** (***) ** ******* ******** performance, ******* * ********** *** ******* made ******** (********** ******* **** *** **-******** ***-***** Camera ******)
**** ****-***** ******* ******** ** ** released ** ***-****, **** *********** ******* a ***** ****** **** ******** ********** for ***** ************, ******* **** **** take * ****** ** ***** ** role *** **** ******* ********** *** will ***** **** *********** ******** ** full ***** *******.
6x *********** **** ******** **********
*********'* ***** **** ******** *** *******'* new ****** *.* ************ *** ***** processing *** ** ***********, ********* ****** language ****** *** ** ***** ****** processor (** ***). ********* ****** * 6x *********** **** *** *********** ***** chipset, **** ** ****-*** ************ *******.
*** ********* ** ******** *********** **** the ****** *********** ****** ******** (***), which *** ********** ************* *** *** many ****** *****. ************, *** ***** offers ** ******* ** *********** **** its ***********
Onboard *** **********
********* **** **** (*** ****** ** ISC ****), **** *** **** ******* runs ****** **** ******'* ****, *** vision ******** ***** ********** ** ***** surveillance ******* ** ********. *** ******* added **** ***** *********** ****** *** more ******* *** *********** ******** ** CNNs.
***** ***** *** ******* *** * dual ****** ** * ****** ****** setup, ***** *** *** *** ******* models **** ****, ** ******* ******** of ****** ************ (***). ******* ********** vision ******** ****** (****)—**** ** ****, Phi-2 *** ****, **** ****** * to *.* ******* **********—*** ****** *** on *** ****, ***** *** ******* up ** * ******* ********* ******** models. **** ****** *** **** ******* and *********** ** ***** ** ********* and ******** ******** ** ****.
********* **** **** **** *********** **** in **** *** ******* ******* *** forensic ******** *** ***** *****.
**'** ***** ***** ** ****** **** CLIP ******* **** **** **** *********** on *** ****. *** *** ****** pretty **** *** ******* ******* **** the ******, *** *** *** **** perform *****-***-**** ********.
Compared ** ******* *** *****-***** *******
***** *********'* **** **** *** ******* VLMs ** *** **** ** ****-**** from *** ****** ******, ** *** process * ******* **********/**********, ** ** uses * ******* *** *****. ** ISC ****, ********* ****** ***** ***** using ******'* **** *****, * ***** that ** * *** *********** ***** than ******'* ******* *******.
******** ** *** ***** ** *****-***** systems, *********** "**-******* ******" ******, *********'* **-****** **** ***** **** limited **********, *** **** ******** *** contextual ********, *** ** *** ********* required. *******, ********* ***** ***** ******* VLMs ** ** ********** ***** ** processing **-***** *******, *** **** *** latency ** ********* *******, ******* *** cloud ******** **** *** * *** larger ********** *** ********** **** *** longer ******.
AI ***** ****** ********** (***) **** ****** *.*
*********’* ****** *.* ************ ********** ** Image ****** ********** (***) ** ******* image ******* ** **** ****, ************ under *********** ********** **** *** ***** and ***. ********* **** **** **** the **** *** **** ***** ******** AI *** ******** ***** ********* *** clarity ******* *********** ***********.
***** ***** *** ******** ** ****** AI *** ***** ********* ***** ** processes, ******** *** ************ ********* ******* compromising ***** ******* ** ********** *****.
********* **** ***** ***** *** ********** between *********** *** ***** "*** ** the **** ***-***** ** *******" *** AI ***:
********* ********** **** ***** *** ***** are ******* ** ******* ******** ********* concurrently, **** ** ** ***, ****** language ******, *** ** ********.
** **********, **** ************** *******'* ** *** ******, ***** **** ********* ***** ********* (line ******** *** *********). *** ****** camera **** *****'* ***** ***, ************* other *** ***-***** ******* ** *** testing.
Lower ***** ***********
*********'* ***** *********** ***** ******** **** **** **, with *** * ** **** ************, while ********* ****** ** *********. ** comparison, **** ******** ******* *****, *** has ***** ** *********, *** ** based ** ** ** **** ************.
** * ****** ** *** ****** architecture’s ************, ******** **** *** ******* technology, *** ***** ******** **** **** 3W ** *****, ******** *** ********** security ******’* ******* ** *********** *** watt.
Adoption ********
********* **** ****, **** ****-***** ******* will ** ******** ****** ** ****, but ****** ******* *********** ************:
** **** *** *** ***** ****-***** cameras ****** ** ****** ** *** next ****** ********. ***** ***** ** limited ************ ************, *** *** ****** is *** *** ****** ******* *** cameras ***** *** *** **** ************.
*** ******* ***** **** ******* **** chipsets ******** ********** ** *** **** five *****, **** **** ******** **** in ***-*** ******* *** ** ******** no ****** ***** ****:
* ******* **** **** ***** **** now, ***'** ***** *** ***-*** ******* using ***** **** *** ***, ****, or ****. ** ****, *** * family **** ******** **** ********** ****** out, *** *** ********** **** ****** be *** *** **** ******, **** our ***** ****** *.* ** ****** that ******** **** ****** ************/***** *** CNNs.
********* *** ** ******** **** ********** on ****-*** ******** *** ****:
****, ** ****** ** ***** ******* up ** *** *** ** **** year *** *** **** ***** ****** CV72 ******* ******** ** *** **********, in *** *** ********** **** **** ramping **. *** ** **** *****, in ****, ****'* ***********, ******* **** of *** *********, **** **** ****** CV72 *******,they **** *** *** *********** *** **** ** ****** *******. *** ** ******* **** ** ***** *****, ***** **** **** *** **** ******, **** *** *** * ******** ******* ** ******* ****** **** ******** ****** ******* ***** **** **** ** ****** ******** ***** ** ****. So we expect that the CV72 at the beginning is really serving our traditional IoT enterprise site, but it will enable new applications in the second half next year. And we expect the lifecycle of this product will be three years, four years, just like before. And also we have CV75 that will ramp up as a mid and low-end product for the CV7 family. So that is the ramping up situation. [emphasis addded]
***/***, **** ****** ********* **** ***** usage ** **** *** ********** ** may *** **** ***** **** ** later.
**** ** * ***** ******, ** fully ********** *** ***** ** ** at *** ****. ** *** **** tremendous ***** ***** *** **** **** bandwidth ***** *** *** ******** ************.