Inspiration
We noticed tons of plastic being dumped into the open, killing the oceans full of life. We want to not only develop technology but also act as a call to action. The purpose is to raise awareness for global warming and pollution, in an attempt to save the environment. With this project, we can give our generation a sense of purpose while also fixing our planet for the future of humanity.
What it does
We use datasets from link to find the general information on ocean currents. Based on the ocean currents, the model predicts where the trash is likely to end up (where the currents end up in a circular motion). We also used link data set for the international ports. The user inputs the data, time, and starting location that they want to find the nearest trash patches. Based in the user input, our application will find the nearest port that the user can start in and the nearest cluster of trash the user can help pick up.
How we built it
Starting off with the datasets, we preprocessed it so the model can compute the data with no problems. Then, due to the Earth being a curve, all the distances we had were calculated using the haversine formula. We detected the accumulation range, which was ocean current accumulation near the port selected, but if we couldn’t find any due to a lack of datasets and processing power, we chose the closest ocean current. We then used 2 models, CNN for spatial data and LSTM for temporal data. We trained it 2 times and used the test data for the third fold. At the end, we concatenated the two models to print out the trash latitude and longitude. We then saved the learning model so we could add it to the webpage. Data is taken from the html webpage, the map and date, and goes through python code running through the model and outputting the port, trash, and distance between each.
Challenges we ran into
One of the first big problems we ran into was that our computers couldn't handle the data we were working with. As a result, we had a lot of crashes and blue screens while training our model for where ocean currents are likely to end up. We had a lot of trouble shrinking our data down and increasing it gradually to the point where we could handle the data being processed. Another problem we had was getting the map to load after the fade. This is because the map didn't know the bounds to be projected so we had to manually project the bounds. Finally, we tried to put markers on the map to where the trash and nearest ports are once they were calculated, but we didn’t have time.
Accomplishments that we're proud of
One of our largest accomplishments is making a functioning machine learning model as freshmen, and managing to connect it to a front end that is easily interactable with the user. Having a smooth connection between the interface and the brains which sends information back and forth, especially using different datatypes such as floats, strings, and APIs like the interactive map, was something we as a team are truly proud of.
What we learned
We learned how to code in the front end, using Python Css and Html together in tandem. We also learned how to successfully code a machine learning model from scratch, and combine 2 models together. None of us had ever used an API before, so using the map API was also a new experience that we learned a lot from. The transition from back end to front, and back, is also something unique that none of us had experienced before. Finally, we learned how to use git and pushing/pulling commands in a repository so that everyone in the team can work together.
What's next for Pluh-stic Clean Up Crew
We had a lot of challenges in finding databases, especially since they could not be accessed without payment or they were private. In our future work, we would like to access these sets that were inaccessible to us so we can improve the accuracy of our model. We also need more resources, like a stronger computer, so we can look further into the future with greater accuracy and have hourly accurate predictions. For ease of the user, we would also like to integrate Google Maps into the website and place markers on the nearest port with resources to clean trash. We also plan on adding ways to communicate with companies or boat owners who plan on doing cleaning expeditions and how much trash they can expect to see in the ocean there.
Log in or sign up for Devpost to join the conversation.