Is Particle OS capable of running an Algorithm for Visual Interpretation?

While I’m waiting for my first hardware to arrive…

I started to review the Official, Public or Verified Libraries — for something that could accomplish this task, or at least get me into striking distance to begin with — then modify an algorithm:

That would run on the Particle OS that could automatically identify Analog Round Dial Gauge Readings using an image captured by a CCD device. The algorithm would probably use polar representation of the dial gauge image to identify the needle position as well as the start position of the dial gauge. Once these are obtained the reading of the dial gauge can be estimated with prior calibration information.

CCD takes a snapshot of the Analog Gauge, Image is processed to find the needle position, needle position is interpreted by using 1/300 degree resolution and based on that value, converts the position to the actual value by referencing a value table – sending the actual value, rather than the needle position.

Depending on the processing power… if the position of the needle is the best that can be done… then the position 1-300 would be returned and further processing could be accomplished on the server side of life.

Thoughts? Advice? Library Entry that I should be looking at?
Slam my head in a door jamb for even thinking this is possible? :wink:

Thanks in Advance!!!
RK

If you had asked about a gen3 device then I would say that @bsatrom is working on porting TensorFlow Lite micro to those devices. The issue with Photon is RAM or lack of. A camera interface is possible (depending upon resolution) but that and categorisation algorithm won’t go together. Another approach is to use a Raspberry Pi with OpenCV or another vision system with analysis and categorisation being done there. I have used an OMRON HVC system - I guess others can suggest other systems - Pixxy?

1 Like

ESP32-Cam modules are dead cheap and also feature more processing power and RAM.

On a Photon you could hardly - if at all - accomodate a bitmap 320x200 @ 8bit as that would require 64K RAM and unpacking JPEG images is no option either - let alone process the image.

3 Likes

These seem ridiculously cheap for what they could do. Have you done any development using Arduino IDE with the ESP32 Cam? I haven’t looked but are there libraries for image capture and analysis?

I have played a bit with these modules but didn’t change a lot in the example code yet. The Arduino ESP32 examples are quite capable already.
So I mainly stuck with the original code only tweaking some settings here and there, adding some debug statements to get a feeling of what’s going on in the code and such.
But that was enough to realise that this is quite some value for the little price tag.

1 Like

Thanks. I think I will buy one. One last question - is there openCV running on it?

Nope, OpenCV is not running on the ESP32.

Hello I am trying to make a face detection system but most of the examples I find are with rasperry, do you currently think this is possible in particle? Thank you very much in advance

@road97, face detection requires to many system resources not available on MCUs. You can, for example, connect the Particle device to a DFRobot Huskylens which can do all the heavy lifting of facial recognition. The Particle device can then act upon the recognition results.

3 Likes