In collaboration with the SNO CORAIL team, the students first listed the different monitoring methods used in French Polynesia (photo-quadrant, PIT) and their advantages/disadvantages. The main challenge: time! The collection of these measurements is extremely time-consuming. Another constraint: the need for an expert in marine biology to conduct the monitoring.
The focus of the student project is to improve current monitoring methods by using Artificial Intelligence to compare images of corals through the use of a convolutional network . The idea is to develop a program that has the capacity to determine whether a coral is alive or dead, on its own. The creation of such a model requires a very large database: "Thanks to the SNO data, which are numerous and already processed, our program is now underway and we will soon be able to carry out comparison tests", Raimana Brasier (Student in charge of the project).
To request SNO data : planes coral cover, fish et temperature univ-perp.fr