The MEGAFAUNA project ambitions to develop a multidisciplinary approach to improve the protection of endangered marine megafauna species. Our approach will combine aerial video-surveys, deep learning, habitat modeling, and marine conservation planning.
The MEGAFAUNA project sets out in the archipelago of New Caledonia, a EU Territory of France that is an extraordinary hotspot for marine megafauna.
The MEGAFAUNA project is funded by a 2-year EU MSCA fellowship. I will be jointly hosted by the MARBEC and LIRMM Laboratories in the University of Montpellier and the ENTROPIE Laboratory in New Caledonia.
Updates on the project's progress in 2020
As many other research projects, MEGAFAUNA was significantly impacted by the covid-19 crisis. Despite the restrictions, I was able to make progress on the detection of marine megafauna in New Caledonia, all of it remotely! For me, the lockdown was an opportunity to learn a new programming language and familiarize with deep learning techniques for automated image processing...
Ultra light airplane of the Air Paradise company fitted with a GoPro camera
Thanks to a collaboration with a local tourism company and funding from Monaco Explorations, we collected over 40 hours of aerial videos of the Poé lagoon, a natural reserve on the West coast of New Caledonia. Visualization of these videos allowed us to spot hundreds of dugongs, dolphins, sharks, rays, and turtles in this lagoon.
Map of megafauna sightings spotted on videos collected in the Poé lagoon
With a great team of students, we localized and identified megafauna species on images - a procedure called 'annotation' which is a crucial prerequisite for applying deep learning techniques. We first focussed on dugongs and eagle rays, two vulnerable megafauna species. We successfully trained deep learning models to automatically detect these species on aerial images, as illustrated below.
Outputs of deep learning detection models
Project's progress in 2021
The first project's article was published in Conservation Biology in August 2021! In this article we provide a deep-learning method to rapidly detect rare megafauna in the field for monitoring and conservation purposes. A press release including lots of beautiful pictures is also available in English and in French.
I was finally able to travel to New Caledonia from May-September 2021 ! There, I conducted an aerial survey of marine megafauna in the western lagoon in collaboration with the local company Air Paradise. The analysis of the amazing dataset collected during this survey is in progress. To see how the survey takes place and discover the marine megafauna species we encountered check out this video!
GoPro camera fitted to the light airplane surveying the lagoon
A selection of marine megafauna species we encountered during the survey
2022 Project updates
Check out our latest publication on vulnerable eagle rays in New Caledonia! Our deep learning model successfully detected 92% of the annotated eagle rays and allowed mapping their distribution across a coral lagoon. Congratulations to our student Lila Desgarnier who did an amazing job leading this work!