In our current society, plastic waste and other articles of trash are major causes for the pollution of our environment. A 2016 study showed that almost 1.25 million metric tons of plastic waste generated in the U.S. was either illegally dumped or littered throughout the environment. This type of pollution wreaks havoc on the beaches, rivers, oceans, and many of the other various natural habitats. Not only is this a danger to our environment, but it is heavily affecting taxpayers. Approximately 11.5 billion is being spent every year on causes to help support better waste management. We want to make sure the world stays clean, so we designed

What it does

We created a mobile accessible web application around the tracking of where people have spotted garbage and a community effort to reduce waste impact. A user can take a picture of litter and the application will post that image according to the location it was taken at onto a map where every user can see. The AI will classify what type of garbage it is by scanning the image it received. The user can additionally decide to create or join a cleaning party to help reduce waste impact along with other community members. Actions will be rewarded with points that can be redeemed for prizes/gifts from vendors.

How we built it was built on a backend based on Python, and Flask, with a Machine Learning layer using TensorFlow. This software uses flask to interact with the server, along with python to interact with various functions throughout the website. A Mask C-RNN was used to analyze and predict the types of trash found in an image. The frontend of the website was done using HTML5, CSS, AnimeJS, and Bootstrap. For map data and points on the map, we used MapBoxGS. This was connected through Javascript to the main website.

Challenges we ran into

  • Implementing mapboxgs and updating trackers
  • Sending and manipulating data from server to application
  • Updating events in response from the data
  • Creating a comfortable ui/ux for mobile users
  • Utilizing a trained model that was based on a older version of Tensorflow to run on current applications
  • Devising a plan to send images from a mobile device to a server to host the machine learning work
  • Coding ML to be efficient and fast to make sure it works in the background with zero supervision or external assistance.

Accomplishments that we're proud of

  • Creating a mobile friendly and accessible app
  • Accurately tracking garbage and updating the data via mapbox
  • Creating a clean and simple ui/ux

What we learned

  • Implementing a clean ui/ux design
  • Utilizing MapboxGS to display garbage trackers
  • Using Mask R-CNN along with taco dataset to classify litter
  • Modifying DOM as data updates

What's next for

We hope that we may be able to obtain a sponsor in order to incentivize people to set up communal meetups in hopes of a cleaner city. Another way that we could potentially improve our app is to implement the use of drones. The process of using drones can be set up in many ways, such as using them in combination with out ML assisted garbage classifying system to locate trash in hard-to-reach places. Eventually, we would want to be able program drones to autonomously pick-up garbage as-well

Built With

Share this project: