TrashOut

Logo

Inspiration

Our inspiration for this project comes from our deep passion for improving our community. Too many times have we seen people just throwing trash on the streets, without caring about its consequences on the environment. Too many times have we seen videos of animals dying and struggling to survive with the huge amounts of trash polluting their habitats. The environment and it's animals shouldn't have to face the consequences of our actions, actions that we can easily prevent. This is why we have created TrashOut, to help reverse what we have done, to help clean up our mess, and to help make sure our planet can have a clean and healthy future.

What it does

The app works by assigning its users a new objective every day. This objective contains different items of litter they should to pick up. They have to take a picture of the trash in order to prove they completed the objective. If they complete the objective, then they receive a certain number of points. As users continue to pick up litter, their points will grow over time. With these points, they can then redeem awards such as gift cards. To ensure people are actually throwing trash away, we require proof of trash bins to be in the picture with the litter.

How we built it

We built this by creating a dataset of trash can images and creating a machine learning model testing to see if trash cans could be identified. In addition, we trained another machine learning using the existing datatset TACO, which includes a variety of images of litter. We then deployed our models to our Swift mobile application.

Challenges we ran into

Our main issue was making sure the model correctly identified a trash can and did not identify other objects as trash cans. We also had to find a way to use our code in our app. Our biggest challenge was converting our python neural network to something that was useable by iOS environment.

Accomplishments that we're proud of

One of our biggest accomplishments is actually making the dataset and also creating successful and accurate models by organizing the images in the dataset we made. We also solved our issue about converting our neural network by converting it to CoreML which can be used by IOS devices.

What we learned

Through this hack, we learned how to effectively troubleshoot issues. In addition, we learned how to build our own datasets.

What's next for TrashOut

We want to increase the images in our datasets to make it as accurate and flawless as possible. We also want to expand and allow people in areas outside of cities or in areas with lower littering to still be able to use the app. This will allow everyone to contribute to this effort and also keep things efficient and effective. In the future, we plan to have ties with companies such as Amazon which can help provide better rewards for our users who have helped clean up our neighborhoods and our environment.