Mr. Bin | Devpost

Inspiration

We are all concerned about the environment and recycling plays a major role in saving it. Waste sorting has been around for decades but people still tend to throw stuff to general waste bin even when it should go in recycling. We wanted to build an IoT solution to the problem, which will help recycling by automatically telling the user where the thing he is holding goes.

What it does

Mr. Bin saves the planet. When approaching Mr Bin with litter the camera will detect the object with a certain accuracy, then it will classify it as recyclable, general waste or non-recyclable. It will finally point to one of the bins below the screen for the user to throw your waste in.

How we built it

We built Mr Bin with Python and Microsoft Computer Vision APIs deep learning algorithms. We also used OpenCV and multithreading to capture the live stream video. After the program recognizes a frame, we used the image tags and confidence levels to go through a data base of possible waste items and then sorted them accordingly. We have also used Microsoft's TextToSpeech API to have Mr Bin saying the destination bin out loud. Finally we wrote a welcoming GUI using tkinter.

Challenges we ran into

At first we had to familiarise with the APIs, but the hardest part was allowing live stream video capture. We achieved this by capturing the video and sending every 10th frame to a seprate thread that will analyse it and display the result, while the main thread keeps capturing and dispatching.

Accomplishments that we're proud of

We had never worked with computer vision before and trying to understand so working with it raised a few problems. The hardest task was indeed allowing live stream computer vision since there wasn't an API for it. We managed it by analysing evey 10th frame of the video capture. We're also very proud of managing to work well in a team that had never work together before.We had to coordinate in order to deal with spaghetti code and finally get the product ready.

What we learned

We have been using computer vision algorithms for the first time. We familiarised with python multithreading and Microsoft's APIs for Congnitive Services and Text-toSpeech.

What's next for Mr Bin

The biggest improvement we want to bring to this IoT solution is some electronic component such that Mr Bin will be a stand alone product which will automatically dispatch the waste in the right bag after detecting it. In this scenario the user just throws waste in the it and the bin will operate automatically by detecting the type of waste and putting it in the right bag using motors and a scale. We believe that this would reduce user effort and improve recyclability like never before.