🐠 About the Challenge

Rising temperatures combined with an increasing concentration of CO2 in the ocean cause the water to be more acidic. This change in temperature and acidity harms animals that build shells, such as corals. Occupying less than 1% of the ocean floor, coral reefs are home to more than 30% of marine life. Apart from the importance of biodiversity, coral reefs provide food, income, and protection for half a billion people on Earth. Predictions are that if the ocean’s temperature rises by 2 degrees, the earth will lose all its coral reefs. 

‍Coral reef surveyors collect benthic images about their marine ecosystem and must store, process, and eventually share this information. Organizations that do regular surveys of their reef tend to take ~100 pictures per survey. When they return to the office, they analyze the captured photographs to infer the health and growth of their reefs, as well as to count and identify fish and substrates and other features important for a holistic and thriving marine ecosystem. The analysis consumes more time than dive - it takes 5 days of analysis for every 4 hour dive.

🎯 GOAL: Creating an ML model for coral detection and classification

A typical reef analysis follows the following steps:

• Step 1: Color correction

• Step 2: Image segmentation of corals (boundary detection)

• Step 3: Identify the coral cover and length of corals (Ruler detection and measurement)

• Step 4: Color & bleaching detection

The goal of this challenge is to automate the first 2 steps by:

• Image segmentation of coral

• Supervised & semi-supervised learning

• Creating a ML pipeline for scientists in the field 

🙌 Results & Final Presentation

Watch our final presentation below:

✨ Team (Colour Enhancement & Coral Classification)

• Parth Dandavate

• Julieta Milan

• Kim van Asten

• Leonard Meyer

• Jaka Cikač

• Harits Abdurrohman

• Nathanya Queby S.