I am sure all of you are aware of the situation we are all in when it comes COVID-19.
Besides dealing with the virus we now have to find a way to get society back in gear and moving yet also not re spread the virus. When business open back up they are going to have to prove that they are making there customers adhere to social distancing requirements.
I think if we put our collective heads together we can come up with a social distancing application for the nano using Deepstream that we can deploy in mass.
Here are just some ideas to start:
It has to be fairly inexpensive so it can be deployed in mass,
It has to have some type of GUI interface to display info.
It has to be able stream so it can be viewed from any mobile device that can log onto a business WIFI.
It has to be able to indicate when two or more consumers are to close.
It has to have some kind of way to show where in a business there are locations where consumers are congregating so the business can find ways to mitigate this.
I believe we already have most of what we need with Deepstream and the Nano.
I have shown that the Nano can run 4 I.P cameras with a POE switch using Deepstream and detect people at 24 FPS.
All the components for this project cost around 200$
Dkreutz
I have downloaded and ran the project on my Jetson Nano. While the app is well thought out and the GUI interface is superb the app runs slow on the nano because of memory limitations. Deepstream on the nano can run 4 I.P. cameras at 24 FPS while tracking people. If we could use deepstream to process the images combined with the GUI from the Smart Social Distancing app I believe we would have a winner.
Great feedback @adventuredaisy! We had this in our roadmap but I did prioritize this request We are going to work on it and probably release Deepstream model in next 2-3 weeks.
See our Kanban board and let me know what you think:
@reza2 It seems to me like such an app could be very useful to enforce workspace safety and prevent contation when people start getting back to work, but less useful deployed at scale in public simply because bluetooth handshakes work really well already to track people effectively and anonymously.
In any case, I’d be happy to help integrate DeepStream into your project. I have some experience with DeepStream in C, C++, Python, and Vala on both aarch64 and amd64. If you’re interested, lmk on GitHub since i responded on the Issue you created.
On the issue of calculating distances. You can utilize a kinect V2 to reference objects and there distance from the camera.
I know the Kinect V2 can really only sense depth to about 15 feet but since the cameras will be in a fixed location you should be able to have enough measurement points to extrapolate distance points farther out.