Quoral

An ML-Based Coral Monitoring System

Comment

Inspiration

Our inspiration for Quoral came from the urgent need to protect coral reefs, which are rapidly declining due to climate change and human impact. Current methods, like satellite imagery, often lack the resolution and real-time data needed to accurately monitor coral health, and manual methods such as scuba diving are labor-intensive and not feasible for continuous observation. We wanted to create an automated, scalable solution that provides high-resolution, real-time data directly from the reefs, empowering researchers and conservationists to make better-informed decisions to protect coral ecosystems.

What it does

Quoral is an automated monitoring system that continuously observes coral reefs and sends real-time alerts on the health status of the corals. It consists of a waterproof device powered by a Raspberry Pi connected to a camera and temperature sensor. The system monitors coral reefs, classifying them as bleached or healthy using AI models. The data is sent via a WebSocket to a FastAPI server hosted on an EC2 instance, which handles requests from the device and our frontend React app. The React app provides an interactive map showing each device’s location, live video feeds, recent images, and detailed health data, helping researchers make informed decisions without the need for frequent diving expeditions.

How we built it

We built Quoral using a combination of hardware and software. The hardware includes a waterproof casing containing a Raspberry Pi, camera, and temperature sensor. We trained an ML model on top of yolov5 to do object detection coral, resulting in a model accuracy of 97%. We then used a K-means algorithm on the detected coral’s colors to classify it as bleached or healthy. Data from the devices is transmitted via WebSocket to a FastAPI server hosted on an EC2 instance, which handles data processing and storage. Our frontend is a React app that visualizes the data, showing aggregated graphs of reef health, interactive maps with device locations, and historical data collected on coral conditions.

Challenges we ran into

We encountered several challenges, including issues with our initial Raspberry Pi camera, which did not work as expected. After some frantic searching, we managed to rent a USB camera just before the library desk closed, saving our project. Designing the waterproof casing and ensuring all the electronics were protected was another major hurdle. We had to protect all the electronics and sensors from any water damage. Balancing hardware and software integration proved time-consuming and complex, as we had limited experience in working with hardware.

Accomplishments that we're proud of

We are proud of successfully combining various tech stacks and integrating hardware with software to create a functioning product. Despite limited experience, we overcame numerous hardware and software challenges to develop a system that automates coral monitoring. The real-time data visualization and interactive map are particularly rewarding features that bring significant value to researchers and conservation efforts.

What we learned

We learned a lot about combining hardware and software, using different tech stacks, and building scalable solutions. This project pushed us beyond our comfort zones, particularly with hardware, and taught us valuable lessons in problem-solving, rapid prototyping, and resourcefulness.

What's next for Quoral

Next, we plan to scale Quoral to multiple devices and reduce the cost of each unit to make the solution more accessible. We also aim to enhance our AI models with more training data in various ocean environments, and explore additional sensors to gather more data points, further helping to protect and preserve coral reefs worldwide.

Share this project:

Updates