Psync GPT-Enabled AI Camera Tested
Large language models like ChatGPT are surging in popularity, with various companies offering products using API calls to enhance their functionality. But how well do large language models and contextual understanding work on cameras for security?
IPVM bought and tested Psync Genie S, examining:
- How accurate are the scene descriptions?
- What objects can it identify?
- Does it miss any people in the scene?
- What was the tested latency on events triggering?
- How accurately can it identify what someone is holding?
- How accurately does it ignore inanimate objects?
- How user-friendly is the app UI?
- How easy is it to setup a context alert?
- How does its pricing compare to Wyze?
Competitive ***** ** ******** *******
***** ***** *** $** *** *** **** ***** and $** *** *** **** ***** and ** ************* ****** ******* *** ~$35**** *** **. ***** ***************, * ~$* ******* ************ ** unlock ********** ************* *** *** **********.
Physical ********
***** ***** * ** * *** camera **** *****-** ***** ***** *** two-way ***** *** ******** ** ***** using ***. ** **** *.**** **** network ************. ** ****** *** ******** characteristics ** *** ****** ** *** video *****.
Executive *******
** *** *******, *** ***** ***** S "*******" ******* ********* ******* *********** issues ******** ******** *** ***********, ********* a ~**-****** ************ ******* ** *******, 50% ** ****** ** *** ***** being ****** ** ******* **********, * 10-second ******** ****** ******* ******, *** general ** ********* ****** **** ** bounding ***** ******* *** ******.
***** *** ***** ****** *** ******, it ***** **** **** ******* ** a *****, ******** **** ******** ****** (e.g., * ****** ******* * ***** near * ***) **** ***** ********* that **** ***** ** ****** *** vehicle. ** *** ******, **** *** be ****** *********** ** ********** ****** improve *** *** **** ** ******** / ** ** *******.
Outlook *** ********** *******
** ********** ***** *********, ******** *********, and *** **** ** ** *******, future ********** ******* *** ****** * useful **** ** ************. *******, *** current ******* *** ************** ** *** calls ***** ****** *** ** ** ~$100 *** ***** ** * ******-******* scene. ********** ******* *** *** ****** to ******* *********** ********* ******* ****.
*******, ** *** *** ** *** decade, ** ** ******* **** **** become * *** **** ****** ******* within ************ ******* **** **** ****** far **** ************* ****** *** **** powerful ********.
Scarce ****** *********** ** *** *******
*** ******* *** **** ****** ****** information ********* **-****, ********* ** ******* on *** *** *** ** ******* of *** ******* ***.
******* * *****, ********* **** ******* founded ** **** *** ***** ** California, ******* **** ~* ********* ********* to ********:
***** ***** ** ** *********, **** Wong, ** ***** ** ************* *******(***** *****), **** ** *** ***** on *** *******'* ********, *** ***** is ** ******* ** **** ****** on *** ******* *******:
*** *******'*** *******, ***** ****, **** **** * ******** *******.
Psync ******* ******
*** "*********.***" ****** ***** ** *** registered ** *** ** ****, *** the **********'* ******* *********** ** ***** as *******.
Shipping ******* ******, **** ************ ** ****
*****'* ******** ******* ********** ************ *** ********, ***** ** *** same ******* **** ***** ******** *** Wyze:
**** ********* ******* ****** ** ***** manufacturer ** * ******** ******* **** (as ***** ** *** *****)******* ****** ********** ** ****** ****'* manufacturer:
No ******* **** *** *******
**** ********* *****, *** *** *** return *** ******** ** *** **** of **********, **/**** **** **, ** will ****** **** ******.
App ********
**** ***** ***** ***** *** ** set ** *** ****** **************** ** **** ** ***** *** overall ***********:
Scenario ************ **** ******** **** ***** *******
** ***** **** *** ***** ******* in *** *****, *** **** ****** the ****** *** ** ******** *** setting *** ******* ** * ****** accurately.
~10 ****** ******* ***** *******
** *** *******, *** ***** ***** S ****** **** **** ** ******* of ~** ******* ** ******* * ViewSay ***** ***** ** ***** ** triggered.
Missed ****** *** ******* ** *** *****
*** ******* ***** **** *** ***** camera *** *** ********** ************* **, ~50% ** *** ****, ** ****** people ** *** *****. ** ***** below, *** ****** ******* * *** and ******* ** *** ***** *** ~6 ******* *** ******:
Small ******* ********* ********** ~**% ** *** ****
******** ******** ***** ******* ******* *** scene, **** ** **** ****** ** tools, **** **** ********* ********** ~**% of *** ****. ***** ** ** example ** * ******'* **** ***** being ******** *** * ***.
False ****** ******* ********
** *** *****, ***** ****** **** spammed ******** ** *** ********* ***** with ** ******, *******, ** ****** to ********** ** *****. ** *** image *****, *** ****** ****** * person ** ******** *** **** *** constantly ***** **** *****.
10 ****** ******** ******* ******
******* ****** **** *** ****** ********* require * **-****** ********, ********* ** complete ****** ** ** ***** ** it ******* ****** **** ******** ****.
40 ********* ***** *** ******
***** *** ***** ****** **** ******* to **** * ******** ***** ***********, ViewSay ****** *** ******* ** ** characters, ******** *** ****** * **** can *** ** ******* *************.
App ********* ****** **** **-****** ***** ***********
** ** "**-******" ***** *********** ** used, ******* **** **** *** ***** with ******** *****, ***** **** ******* the ***** ***** ** *** *****.
Only **-****** ***** *********
*** ***** ****** **** *** ***** constant *********. ******, ** ******* ****** events ** **-****** ********. ***** **** approach ******* ** ****** ****** *** being ******** *** ** *** ** knowing ****** **** ******.
Versions ****
*** ********* ******** **** **** ****** testing.
- ***** ***** *: *.**.*.****
**** ** * *********** **** ** the ****** ** ***** ************, ****** this ******* **** *** **** ***** of ***** **********.
**** ********* **** *** *** ******* logistical ******** ** ********** ****, ** will **** *********** ***** ** ********** a **** ************* ************* ** **** is *********.
** *** **** *******, ** ********** is **** ********* *** **** *** their ********** *** ********* *** ****** succeed.