Redsea-Coral

Dammam | Leveraging AI/ML for plastic marine debris

Leveraging AI/ML to detect plastic debris in the ocean

High-Level Project Summary

We studied the problem of plastic debris in marine environments using NASA literature . We found that this problem is highly correlated with bad weather and coastal populated areas. We developed a working end-to end solution to find and determine these areas of high risk of debris, scan satellite and public images (social media,/news/public cameras such as beaches) for potential debris in such areas using machine learning. The proposed solution is fully automated, utilizing all available data and eliminating the need to manually report and track debris. The solution will follow weather conditions in such areas and apply machine learning on such areas to determine any migrations of debris.

Link to Project "Demo"

https://docs.google.com/presentation/d/e/2PACX-1vTf3cCBqK4cSPt4zeBI1NWoBn-4HdLIITIfMesEOG1boOyUO0muGCV-yuOZ9udPXw/pub?start=true&loop=false&delayms=60000

Link to Final Project

https://redsea-coral.anvil.app/

Detailed Project Description

We started our project by studying the published resources related to the challenge. We found that plastic debris are highly correlated with bad weather and coastal areas. With the number of satellite systems observing the globe on daily biases, processing images for the full globe is computationally impossible. The team developed a method to identify areas of interest with high chance of debris generation and migration which are mostly coastal areas, river streams and seaports. By focusing on such areas, we eliminate the need to process 90% of the satellite images.

Augmenting such areas with NASA satellite and weather data, the task became more focused on areas with high risk of plastic debris. Using satellite images, news data and social media, the team developed an end-to-end solution from image collections, process to plastic detection in such images.

This is done by using Machine Learning and Convolutional Neural Networks (CNN) to detect plastic debris in images and compare how specific areas change compared to before and after a certain period of time through satellite image history. Then tag them with geospatial and time metadata to make such data available in a database for communities and government to

action their removal.

During this hackathon, we have developed a working proof-of concept how all the pieces fit together from potential debris area to areas were debris are actually detected. The machine learning model developed for this purpose showed accurate result for images, recorded media and camera feeds. This make such solution possible for all kind of information sources whether it is satellite images or public cameras in places like beaches.

The developed solutions has the following functionalities:

Get live relevant bad weather events from NASA and show location on map
Show all monitored area on map
Get recent and previous satellite images of selected event location from NASA
Get social media images
Detect plastic waste in images and video

Introduction video: https://youtu.be/XEnHWxSIVZo

Sample Detection Video: https://drive.google.com/drive/folders/1KYRJlPvPQX3rFHQq4rxdfrFOMhUpiT5V?usp=sharing

slides: https://docs.google.com/presentation/d/1k6mW5j0LuBRRovXFbPggyVv-n2EAVqpP/edit?usp=sharing&ouid=105218294973775459489&rtpof=true&sd=true

Space Agency Data

Multiple sources of data were used to solve this challenge. For space agency data, we used NASA literature, satellite images and weather data.

We used NASA literature to study the problem and understand its key factors. We found the problem is highly correlated with bad weather and coastal areas.

We used the weather data to determine areas where we need to look for potential debris and debris migration. Such high severe weather conditions, rain, high wind, and cyclones. Once such events are identified, plastic debris detection is applied and historical satellite images from NASA are used to compare how the area has changed before and after the weather event.

Hackathon Journey

The team experience with Space Apps is great. We learned many things about how space data is collected, processed and used in many of our daily lives. We chose this challenge because we all live in a costal city and we understand the impact of plastic material on the cost and the marine live.

We approach this project with a plan on how to we want to address it and distributed the work between the teams. We were always in contact to support each other and ensure continuous progress. This gave us the opportunity to mitigate any issues early on before the set us back.