Analysis of an image classifier created by MVNCS

[idata]

Goodmorning

I'm creating an image classifier using Raspberry Pi3 and a MovidiusNCS.

For the first steps, i've largely followed this guide: https://movidius.github.io/blog/ncs-image-classifier/

After i've created (and fixed) the program in Python3, i wanted to push myself over:

• automatic classification of a small custom database of images (done)
• read and analyze memory usage (RAM and CPU) or even more stats from MovidiusNCS and Raspberry to confront them
• create a new custom net dedicated to do security and video recording
• doing detect+tracking using my Raspberry+MovidiusNCS

At the moment i'm blocked on the second step: there are some build-in commands or Python3 (using mvnc library) code lines to read and register stats from the MVNCS?

You know some other way to do this thing?

Actually i'm using standard memory usage function like this ones but i don't think them can solve my problem…basically because i need to distinguish the work from Raspberry from the one done by MVNCS:

CPU_Pct=str(round(float(os.popen('''grep 'cpu ' /proc/stat | awk '{usage=($2+$4)*100/($2+$4+$5)} END {print usage }' ''').readline()),2)) tot_m, used_m, free_m = map(int, os.popen('free -t -m').readlines()[-1].split()[1:])

Sorry for my english and explanations, i'm a real newbie on this things

Thanks for help and support

[idata]

Updates:

I've installed the NCSDKv2 so i could get, thanks to API, some more information (like the Movidius temperature)

I've installed OpenCV and the NCSDK examples……now i'm trying to run and operate with the TOOLs like mvNCProfile

Someone have try to run and extract information from this TOOL?

Thanks

[idata]

Updates:

Actually i'm at the last chapter of my work: i'll use yoloNCS (https://github.com/gudovskiy/yoloNCS) to do real time object detection

I'm gonna use this guide: https://www.pyimagesearch.com/2018/02/19/real-time-object-detection-on-the-raspberry-pi-with-the-movidius-ncs/

If someone has something to recommend, i'll read asap

Thanks

[idata]

I've literally finished the last step by creating an object detection program using MobileNet-SSD (not yoloNCS)

If you know some way to analyze other performance or similar on NCSDK, please comment below

Thanks

[idata]

@MattiSpark you can get memory usage on v2 of the API (new api)

Look for option RO_CURRENT_MEMORY_USED in: https://movidius.github.io/ncsdk/ncapi/ncapi2/py_api/DeviceOption.html

There's some example code for options here