Self-supervised representation learning from multi-domain data for land cover change detection

- Implementation of a multitask learning method on pytorch lightning for change de detection and characterization- Use of a Bayesian approach to get the model uncertainty on its predictions (Monte Carlo Drop Out)

Continuation of the internship as the final year school project. - Training of the cGAN model in the grid5000 cluster. - Assessment of the cGAN ability to reproduce correctly changes

Fusion of Sentinel 1 and 2 with Conditional Generative Adversarial Networks. The project aims at simulating Sentinel 2 images in bad weather conditions using radar data. I have worked in the data preparation (preprocessing with ESA SNAP software), models implementation (Tensorflow) as well as training on HPC.

Satellite image processing, object detection on SAR imagery. Object classification using both deep learning algorithm (SqueezeNet, VGG) and decision tree algorithms (Random Forest, Gradient Boosting Tree)

A project which consists to find a correlation between brain signals. Classification of brain signals using machine learning methods