OpenDataCam: An open source tool to quantify the world

Dear Jetson Community.

We’d like to show you a project we’ve been working on. It’s an Image Recognition project that runs on all Jetsons and lets you count and track moving things. We call it

OpenDataCam
An open source tool to quantify the world

this is a short intro snippet
https://vimeo.com/351143472

From our website https://opendata.cam you can find all the intro texts like the latter:
‘OpenDataCam’ is a tool that helps to quantify the world. It’s here for everyone to use! With computer vision OpenDataCam understands and quantifies moving objects. The simple setup allows everybody to become an urban data miner.

for more in-depth info refer to the github repo: https://github.com/opendatacam/opendatacam

A quick run through of the project:

We attached a webcam to a Jetson board wich is running yolo including a tracking software we built. You can access the system via a smart device that let’s you define areas to track and count and once you’re finished also export data. The cool thing is that you can use this system for surveying without saving video data thus not intruding data privacy of counted objects. And also that it’s completely open source and transparent where data goes and what happens during the counting algo.

Also don’t miss the walkthrough of the interface:
https://vimeo.com/346340651

If you want to go really deep also check out the tracker we built: https://github.com/opendatacam/node-moving-things-tracker
also a blog post on our website to that: https://www.move-lab.com/blog/tracking-things-in-object-detection-videos

We hope to contribute to the citizen science community with this device and encourage people who are interested in mobility and planning to make more educated and data driven ideas become reality, especially for transport modes that don’t state of the art counting methodology and lobby yet (for example cyclists, pedestrians and the like).

We already have some real world examples of where ODC is being used, check our website for details on that and also the suggested “shopping list” if you want to make your own.

move lab is a creative working environment working on prototypes, ideas and scenarios around the future of transportation. We initiated this project amongst others and hope to find people who are just as interested in mobility as we are. Our website can be found here: https://move-lab.com

Happy Counting :)
Raphael form move lab

That’s great

I already tried on my jetson nano, it works at 12-15 FPS with 5v 4A of power, it works very well.

Thanks for sharing.

Thanks for letting us know - happy that you got it running. Just curious, what have you counted how did you set it up? If you ever have any ideas or problems you stumble upon please let us know on GH: https://github.com/opendatacam/opendatacam

I tested counting persons in the offices, to know which corridor is more used.

But for some reason the RAM becomes full after about 20 minutes, I think this is relate with the darknet, because I tested with docker and without docker and get the same result.

Yes, this is a known bug. Thanks for reporting, unforetunately there is a memory leak. --> https://github.com/opendatacam/opendatacam/issues/131