Research internship on deep learning for remote sensing, jointly supervised by LETG Brest and IRISA Vannes
While coastal cliffs represent about 52% of the global coastline, cliff erosion is likely to increase as sea level rises, endangering surrounding populations and infrastructures. Bringing new challenges to research, climate change calls for the development of innovative methods for Earth observation and monitoring.
This research topic is part of the HIRACLES (CNES) project (https://www-iuem.univ-brest.fr/pops/projects/hiracles?jump=welcome) which aims to develop a new optimized approach to detect and quantify cliff front erosion using Pléiades imagery.
The master thesis will be carried out in connection with the CICERO doctoral thesis (Contribution of multi-angular spatial imagery to the monitoring and understanding of cliff erosion), started in 2021.
The main objective of this master thesis is to develop a methodology to provide a coarse delineation of rockfalls in a dataset made of oblique images. The originality comes from the fact it will be achieved without using multi-date image pairs comparison (i.e. no siamese neural networks). Several detection scales can be tested (leading to patch-based, pixel-based or object-based approaches).
(c) GdR 720 ISIS - CNRS - 2011-2022.