Greenseer | Devpost

An IOT device for increasing recycling awareness
Our hack
SQL server
Mobile app

Inspiration

Recycling is a complex and underappreciated thing. People constantly mix up bins intentionally and accidentally, and this leads to the detriment of everyone. Our team wanted to come up with a solution to this problem by alerting people who do not recycle properly, but instead throw in trash or other non-recyclable materials. We hope that by allowing communities to see how their doing and associating bad noise with not recycling properly, we can help improve recycling as a whole.

What it does

It launches the GoPro through a wireless connection to record video, sends that data back and then converts the video into images. The images are then used by being put through a machine learning algorithm to determine whether or whether not the object that went into the recycling bin was supposed to, and then plays an annoying beeping noise if someone throws and incorrect object in. The data from the machine learning model is then sent into Google Cloud into a MySQL database where we query the data and then send it out for receiving. A mobile application is then running which sends requests for the data to the Google Cloud MySQL database, updates accordingly and provides a visual representation for seeing the data.

How we built it

We used the goprocam module on Python to remotely hook up to a GoPro Hero 5 which we borrowed from the university library. Then we converted the video files received from that into images and trained our machine learning module on a good dataset for our project. Then we set up Google Cloud SQL to store our data and create a place for the data to exist to then get sent to our application. The mobile app then issues a request constantly and receives responses determining whether it should be updated or not.

Challenges we ran into

GoPro requires you to have your wireless connection on a computer be connected to them, which caused stress in terms of how to get the data off the camera and onto WiFi. Google Cloud SQL was also difficult to receive requests from, burning a great deal of time to implement and then utilize fully.

Accomplishments that we're proud of

Connecting to the GoPro virtually was extremely cool, especially because it is a device that cannot go on the internet by itself. Getting a properly working machine learning model and online SQL database were also very satisfying and rewarding.

What we learned

To cut down on scope when we have to - for this hackathon we implemented an extremely high number of different things which is great, but sometimes we wasted a great deal of time on something that could have been done easier had we gotten more comfortable with a few devices over the years and used those.

What's next for Greenseer

Finishing the model and trying to deploy them out to a community when we have the chance. It would be interesting to see whether this positively impacted our communities.