Simple one or zero effort?

Expect something unexpected ;)

Comment 1

-- The TRY-OUT Links require authentication (as there is just the production system and no testing instances to use). The second links is to the grafana where you could get a first idea about the data available. Please reach out to Ben and he will show you these systems whenever you like 👍

Inspiration

A member from us is in the "open source smart home scene" active, with some small own projects and has a passion for this topic. In the last years a lot of experience regarding the data in a typical (non-rich!) home/flat. Ben's own "smart home" is up since about three years already and generates a bunch of data every day. He also does some "reasoning" from the data already to get additional value.

A lovely example that makes washing the laundry, much more convenient (due to the building we live in, we cannot hear or see when its finished and such things...) is that depending on the current energy consumption values (acquired by a TP-Link HS110 smart-plug between the washing machine and the wall outlet) I can reliably detect the state in which the w. machine or tumble dryer are. This information prevents from useless journeys down in the basement and also is a handy reminder (every XXmin) to get the laundry out of the finished machine.

Besides that, a sad personal connection to this topic also pushed me to finally get the chance do something, at least what we can do here and now, against this awful disease or illness or whatever word is used for cancer....

What it does

We change the actions. We do not just ASK the user but we also OBSERVE him on our own, leading to unbiased data! And, for that stage of our system, there are no direct privacy violations to fear. First because the data is generated and processed on a single system at Bens home in Stuttgart. Nothing is exchanged with someone else. A "further step" for this could be to also utilize cameras! Ben uses them to observe his cat, checks if she brings home some of her mice friends, tracks the time outside/inside etc.... For the cancer case, we could imagine that there could be very valuable data emerge when using some pose estimation algorithms that can detect if one repeatedly touches his back or knee or something and then to further conclusion.

How we built it

There wasn't that much to build (Alexa, Google Home and such do not like the Hackathon Networks :/) but "the typical way" we would say. VS Code Insiders and a Terminal is almost everything needed :) As mentioned before, the (open source) smart home system used, Home Assistant (https://www.home-assistant.io) was already there, inclusively the data from the last YEARS

Challenges we ran into

I, unfortunately, couldn't convince my team that this solution has much more value than a "UI-relaunch" of the current questionnaire. We split friendly, but yes, one expects something else when traveling to a hackathon... Besides that, we would need someone like, for example, you Roche, to do the next step, see next chapter :)

Accomplishments that we're proud of

I am confident that the basic principle is working, in production, since years! But probably its a little easier to do this with my home than with real people. But - I am also sure that simple things even could directly be used in a "Questionaire System" like proposed here. For example the number of fridge-door openings per day, how often one is on the toilette, even how long... But also the duration of sleep, the time one goes to bed... By correlating all this data and comparing it with historic values from last week or things like that we can detect patterns. Anomaly-Detection algorithms could be used to detect, well, anomalies in our data (patterns)...

But to be fully clear, to evaluate which data has to processed in which way to answer a specific question in the questionnaire is far beyond our capabilities, due to the lack of skills in medical research and actually doing a study in real world... But we think now it's your turn again! We are sure you would have everything needed to investigate this further or could at least provide it! :)

What we learned

We are not sure yet what we learned 😅 what we did right/wrong yet... Lets wait for the next few hours and see what happens

What's next for Simple one or zero effort?

So, lets make a deal! We proposed the ideas and are convinced they WILL work! Now we are giving the ball back to you. If you like one of the ideas, please validate/evaluate it with an own prototype or whatever methods you would use and do whatever Roche does to decide if something is worth working on it or even really implement it in production... In the meantime we will also do further investigations and diving deep into our data lake to find even more hidden and more complex complex information-treasures.

And then, we meet again to HackZurich 2022 and talk about the next step in our project 😉

Deal? 🤝

Built With

Share this project:

Updates