Enviro streamlines the recycling process by providing users with all the information that they need without any of the hassle. It (1) classifies the user's image based on what kind of trash it is, (2) teach the user how to recycle that item, and (3) measure the user's impact on the environment.
A lot of members on our team care deeply about the environment and wanted to make strides in improving the current state of pollution and waste management.
What it does
The user takes a picture of a piece of trash and our product classifies that trash based on 14 categories. Based on that category, the user is taught how to recycle, repurpose, or dispose of that trash. Moreover, the impact of the user's actions is measured and quantified.
How we built it
We developed a computer vision algorithm using Google Cloud Platform's AutoML, and ended up with a model that had around 95% accuracy in detecting and classifying trash across 14 categories. We also have a web app created using React and Redux that educates the user on how to properly dispose of their trash and measures the impact of the user's actions. Amy and Pooja worked on image recognition and research. Amy was also the data engineer. Isabel worked on the UX/UI and front-end. Richany was full-stack as well as the architect.
Challenges we ran into
Our greatest initial challenge was the refinement of our idea. It took us around half the allotted time to finalize what project we would make and how we would make it. This led us to have less time building than everyone else. Moreover, since the team has barely seen Google Cloud Platform, we had a steep learning curve, especially with AutoML. We had the most trouble training the model against hundreds of images and deploying the model on the cloud.
Accomplishments that we're proud of
Despite meeting each other for the first time during the hackathon, the team learned to work really well with each other and became good friends. We are very proud of all the new technologies we learned
What we learned
That software friends are the best friends <3 We all learned how to properly utilize Google Cloud Platform's Computer Vision API and GCP in general.
What's next for Enviro
We took scalability into account for our project, so that it would be significantly easy to include more features later on. Some features the team has suggested include geolocation, to recommend exactly which facility they should dispose of the item to.