DEV Community

Cover image for Real-Time and Still Image Classification

Real-Time and Still Image Classification

Rose Day on August 31, 2018

In the past few months I have been working on utilizing the Intel® Movidius™ Neural Compute Stick (NCS) to do some rapid prototyping with deep neur...
Collapse
 
ben profile image
Ben Halpern

It’s amazing how much more accessible this stuff has become.

Collapse
 
rosejcday profile image
Rose Day

Agreed! It has been very interesting to learn about and get into. Amazing what can be done with it.

Collapse
 
ben profile image
Ben Halpern

A lot of this post still goes over my head though 😄

Thread Thread
 
rosejcday profile image
Rose Day

I feel the same still at times. It has been a lot to absorb. Very interesting but deep topic to get into. I have been learning much as I go along but there is still so much more to learn!

Collapse
 
harishkgarg profile image
Harish Garg

How is it performance wise? What kind of specs are for your setup?

Collapse
 
rosejcday profile image
Rose Day

Hello! I have an Intel Nuc that contains an Intel Core i5 running Ubuntu (64 bit) 16.04 with a 1 TB drive. With this, I have been using the Intel® Movidius™ Neural Compute Stick and a GeChic touchscreen. Performance wise, I am not sure what I can really say for this. For still images, it takes a few seconds to run when I have only a few images. I have not run it on a massive dataset yet. With a live camera feed it is relatively quick, second or so, to detect a person and save a screen shot. The only issue I had was with too much light pollution, which made it hard for the camera to realize a person was present.