A Bio-inspired Collision Detector

[patrickpoirier51]

This has just been published:
https://scirate.com/arxiv/1801.04530

Its a a novel bio-inspired collision detector and its application on a robot. The detector is inspired from LGMD neurons in the locusts, and modeled into an STM32F407 MCU. Compared to other collision detecting methods, we focused on enhancing the collision selectivity in a bio-inspired way that can considerably increase the computing efficiency during an obstacle detecting task even in complex dynamic environment.

That would be a great add-on to the OpenMv libraries.

[kwagyeman]

Hi, for the next release about to come out I finally have optical flow working really well.

[patrickpoirier51]

Cool, I saw you progress on this and will certainly give it a try.
Did you looked at the paper above?
We dont see much neural network implementation on STM, could be interesting

[kwagyeman]

That algorithm is pretty easy implement. Would you like to give it a shot in C? It seems to require 3 frames in RAM to work but you can easily store that now with the new alloc_extra_fb command.

[patrickpoirier51]

Yes I can give it a try in C but I would need some guidance from you

Btw this is a more technical paper about the system
http://eprints.lincoln.ac.uk/25279/1/20 ... Bfinal.pdf

[kwagyeman]

If I wrote the input/output part and left you to just write the raw algorithm would that work?

I can make a branch and put the input and output stuff together pretty easily.

A question though, how do you see a img method working? What arguments should it take... what kind of object should it return.

[patrickpoirier51]

OK I have found a trace of the code here: http://ilab.usc.edu/cgi-bin/secure/view ... ots/LoBot/
I will extract on my github , to test and I will ge back to you

[kwagyeman]

If you have code it's way easier for me to port. A lot less of my time then.