XSGJ | Devpost

object detection on test image
recommended recycling bins within 500m of postal code
recommended recycling bins within 500m of postal code

Inspiration

Based on research done in Singapore, 40% of what people throw into recycling bins cannot be recycled. Surveys have also shown that 45% of the people are unaware of what can and cannot be recycled and 42% of people find recycling very inconvenient. These findings tie in closely with the given problem statement, which spur us to design a platform to not only educate the public on what items are recyclable, but also for people who want to be part of this movement to become better informed on the availability of recycling bins in the vicinity.

What it does

Our webapp helps to increase recycling rates via a three-pronged approach. Firstly, the app suggests the recycling bins within 500m from the user's current location. Secondly, the app helps to classify an image of items into the respective recyclable categories, namely plastic, metal and cardboard. Lastly, there is an incentive system implemented such that users are able to earn points proportional to the weight of their recycled items, which in turn could be used to redeem other items.

How we built it

We used React JavaScript library for the frontend and Django framework and SQLite3 database for server-side development. We used the PyTorch framework to train a model on the ZeroWaste Dataset for object detection of recyclable items.

Challenges we ran into

We had trouble acquiring readily available datasets for recycling bins dispersed across Singapore, therefore we substitute it with bus stops postal codes and coordinates instead. Nonetheless, this could be an apt opportunity for us to realize the wide outreach and potential that could be achieved by situating a recycling bin at each bus stop. Furthermore it will be easier for people to remember where the recycling bins are if they were to be placed at every bus stop. Besides, we also had trouble deploying our backend server onto the cloud due to the large size of the PyTorch library.

Accomplishments that we're proud of

We are proud to be able to code out a minimum viable product within less than 24 hours using very versatile tech stacks.

What we learned

We managed to hone our technical skills within such a short period of time. We also learnt to work in groups better. Due to the complexity of our application, we learnt to communicate with each other more effectively and convey our thoughts in a manner that is well understood.