Our inspiration was to focus on a meaningful solution to environmental sustainability. While we knew that we can’t solve all of our environmental issues, we wanted to design something that would encourage people around the world to clean the Earth, our home.

What it does

Our project takes user input in order to create a global database, which can map the trash around the world. This creates a unique opportunity for global reforms targeted at preventing large scale littering, which threatens our environment and the very world in which we live. Additionally, our app allows us to quickly get access to large amounts of recyclable material around the world, which has numerous ways to be repurposed and sold while simultaneously helping the environment. For example, multiple companies already buy aluminum cans, cardboard boxes, and various other pieces of trash. Outside of the US, countries like India have little known economic systems built upon trash and recycling where similar projects have attempted to locate the trash. However, our project uniquely maps the trash globally and verifies the collected trash through computer vision and ML systems.

How we built it

We utilized a variety of languages and apis including: html, css, javascript, python, sql, google-maps api, and google-cloud api. We needed to use concepts from both machine learning and computer vision in order to have achieved our goal.

Challenges we ran into

Our primary challenge was trying to integrate the machine learning program written in python into our html documents. Specifically, we struggled to use the SSIM from the python cv2 package. We attempted to use php to load and access the images through a database, but we quickly discovered that php databases struggle to store images. We also had some initial struggles connecting to the CloudSQL database, which we eventually overcame. Unfortunately in the end we ran out of time to integrate the machine learning algorithm into the website.

Accomplishments that we're proud of

While we had some struggles, these struggles lead to numerous accomplishments. For example, while our ML model couldn’t be integrated into our website, we were able to train our model even with limited data to have a very high accuracy in testing. Additionally, we were able to create a database for our locations where we log our trash, which we successfully mapped. We are also proud of our login system, which allows users to create accounts and log their trash collected.

What we learned

This hackathon taught us a variety of things, both coding applications and morals. We learned how to run a php server to start a local web host, implement machine learning algorithms, and utilize various apis. On the other hand, we also learned we can get a substantial amount of work done by splitting up tasks even though it can be difficult to integrate it in the end. Despite this, we still learned about how to integrate python and html though we didn’t have time to implement it.

What's next for TrashTrackers

In the future TrashTrackers has multiple goals. The first and foremost would be merging our ML program into the database. Next we would make sure to get a larger training data in order to increase our accuracy. As a team, we also have the goal to make our project more wide scale in order to make the global database work well. This means we plan to use various platforms in order to spread knowledge about our platform. Given that we could get a large amount of testing data, we also would like to use computer vision and a neural network in order to classify the type of trash (eg. plastic, aluminum, cardboard, etc.) in the picture and even classify the type of plastic.

Share this project: