Hi there, just received my OAK1 few hours ago, installation and first tests went incredibly smooth! This little thing is amazing! My goal is to improve a solar-powered, stand alone prototype sensor that can count cars, bicycles and pedestrians passing on a street. I already do (badly) that with: - a Canon surveillance camera; - an Arduino with a memory card to run the openCV script that classifies vehicles and store the counts; - a SIM card to send data (only counts, not videos). Can you point me towards some useful tutorials to approach the problem? Also: would you suggest some particular modules/products of yours to add GPRS transmission capabilities to the OAK1? Thank you so much for this amazing product! We are 100% with you in improving cyclist safety, you may be interested in this project of our: http://www.maps.wecity.it/en

Thanks @"dhruvsheth-ai"#p598 for pointing me in this direction, I will surely give it a look!

How to start to count cars, bicycles and pedestrians

ferrix

Hi there,

just received my OAK1 few hours ago, installation and first tests went incredibly smooth! This little thing is amazing!

My goal is to improve a solar-powered, stand alone prototype sensor that can count cars, bicycles and pedestrians passing on a street. I already do (badly) that with:

a Canon surveillance camera;
an Arduino with a memory card to run the openCV script that classifies vehicles and store the counts;
a SIM card to send data (only counts, not videos).

Can you point me towards some useful tutorials to approach the problem?
Also: would you suggest some particular modules/products of yours to add GPRS transmission capabilities to the OAK1?

Thank you so much for this amazing product! We are 100% with you in improving cyclist safety, you may be interested in this project of our: http://www.maps.wecity.it/en

dhruvsheth-ai

Hi, I'm not sure of an implemented example out of the box with the use case in mind, but in case you wish to implement an in-depth monitoring of traffic, You could have a look at the Algorithms used here. It might need a lot of optimisation, but might help:
https://github.com/alex-drake/OpenCV-Traffic-Counter

Brandon

Thanks dhruvsheth-ai and also sorry about the delay @ferrix - been swamped with logistics stuff these past weeks.

Thank you for the kind words, and I'm glad it worked out of the box well for you. And super cool on http://www.maps.wecity.it/en. I'd love to make a bike-mounted solution that reports to your system for real-time risk assessment and reporting based on position/time/day/condition/etc.

In terms of an example for what you are building here, we have a start, but it's a bit initial currently, and we will be making it better using our Gen2 system, once Gen2 is a bit better/further along:

https://github.com/luxonis/depthai-experiments#people-tracker-here

This application counts how many people went upwards / downwards / leftwards / rightwards in the video stream, allowing you to receive an information about how many people went into a room or went through a corridor.

So actually if you change the model used, and then the logic to count per class (car, bicycle, and person), you should be able to use it as a basis for your counter.

Thoughts?

Thanks,
Brandon

ferrix

Thanks dhruvsheth-ai for pointing me in this direction, I will surely give it a look!

ferrix

Hi Brandon, thanks a lot for this suggestion! I am, too, stuck in too many things, hope I could focus more in this part of the job! The application you point out might be a very good start for us, I will check it out and come back with a feedback.

It would be super cool to add your solution to our system, let's talk about this when you see it fit!
Thanks again, and keep up with the awesome job your are doing!

Brandon

ferrix Awesome, thanks! And likely not too long from now we'll have a more sophisticated version of that based on our version 2 software instead of version 1 (which is what that is based around).

Thanks,
Brandon