Very cool man! Great project, and an excellent demo!

I have to say, the fall detection is great and it’s easy to see how useful this could be in a tonne of different applications. Even as part of a standard CCTV system in any building; it might be primarily for security, but why not also add this function so it could help people too? Any store, shopping mall, workplace could benefit from that.

But most of all, I’m really impressed with how it doesn’t trigger the alarm! Being able to sit, kneel, etc without false positives is outstanding work!

Thank you Spanner, and thank you for the opportunity to participate!

Hi ​@moorebrett0 ,

Congratulations on your final submission!
I really liked the way you’ve used different rules and guards to avoid false alarms as much as using the pose information as is - it comes close to scene analysis which then has a broad range of applications. And as you’ve articulated, I couldn’t agree more that your project design provides an “easy path to real world pilots” with an off the shelf camera, $20 smart plug, and cost and energy efficient Metis!

Thanks a lot for your engagement and good luck with future projects!

Best,

Radhika

​@Radhika J - thank you so much for your kind words.  I thoroughly enjoyed building this and have already begun working on my next feature-set and rules in a separate branch in my github. I’ll be posting updates here and helping out others in the community for sure!

Thank you for sharing your project Demo and other instruments details. I think I will press "gold button" similar like in the Talent people TV program. This is amazing project and should be used widely everywhere. You didn't won 1000 Pounds or Dollars, you have won millions. Congratulations. Your project will get real investments from serious Companies. Why I think like that?. Because has very wide opportunities to be fit in many environments. Safety is very important in many environments. To be honest where I was travelling in some countries there is no safety at all like Thailand or other countries and they are very small statistics with injuries. Why?. Because human intelligence is very complicated and very advanced to understand many processes involved with our behaviours. But your projects is for these bodies who cannot control themselves, who need additional support or are very ill. Your project can also be used for other group of people not necessary ill or disabled or elderly, can be used for everyone. As a simple example many people think they are very fit, eating well, doing some sport no drinking alcohol or using some drugs. Yes they can be affected by heart attack or other unexpected issues like gas leakage in the property. They have only few minutes or seconds sometimes to survive and your project can do it. send alert to emergency people. People doing jokes and it can be a joke but sometimes the quick reaction will save their lives. Congratulations to you. I have only one technical question: "How did you managed with space on Aetina?. This is just demo or you can use the system data all the time without managing space on the memory, or you redirecting system data not to be stored on the Aetina?” . Thank you.  

Hey ​@Mariusz,  right now storage is super simple, it’s real time inference processing the frames in RAM, I don’t store any video on device. So it’s basically the model, 10mb of JSON metadata (that gets rotated out) and optional logs for when I was debugging. 

Thanks for your vote of confidence, I think this project definitely has the ability to help people all around the world, I’m hoping to find a mentor in the space that can help take it from prototype to a reality. 

Thank you moorebrett0 for the reply. I understand. In my case system files was very limited with space because system consumed space for itself data. This is why I asked. Yes this project can be used in any computer unit compatible with Metis. That is great. I asked for storage for other reason also. Sometimes is needed to store some data for security or other purposes. Nice design and is worth to spent more time on it, expand and investments.   

Hi ​@Mariusz If your concern is about the lack of disk space on the Aetina board, then you can follow what some of us did when we set up our boards. You can add an SD-card or M.2 SSD and use links to map folders so that most of the data can stay on the SSD/SD-card instead of consuming the very limited space on the flash. 

I documented this procedure so you can follow along if you like:

Full procedure to re-image and prepare an Aetina RK3588 board with Metis M.2 | Community

Hi ​@moorebrett0 , great project, congrats.
I noticed you are classifying fallen / non-fallen states using a set of rules. This is a prime candidate for machine learning. Here is an example: https://github.com/PIC4SeR/AcT