Deep Learning for Image Analysis @ AgroParisTech

Welcome to the lectures and tutorials on deep learning for image analysis organized at AgroParisTech as part of the IODAA master’s program! This page is under construction. It will host the lectures and tutorials given in January 2026, following the schedule below. As this is the first time we present this course, we expect some errors, bugs and issues. Please, send feedback at romain dot thoreau at agroparistech dot fr.

Schedule (First edition, 2026)

Date	Lectures / Tutorials
Friday, 16. January 2026, 09:00–12:30	Session 1: Classification of Satellite Image Time Series with Transformers
Tuesday, 20. January 2026, 14:00–17:30	Session 2: Deep Learning for MRI reconstruction: about instabilities and hallucinations
Friday, 23. January 2026, 09:00–12:30	Session 3: Self-supervised Learning of Visual Representations

Lectures

• A brief introduction to remote sensing, optical satellite images, and tutorial 1

• A brief introduction to Magnetic Resonance Imaging (MRI) reconstruction, and tutorial 2

• A brief introduction to the self-supervised learning of visual representations, and tutorial 3

Tutorials

• Classification of Satellite Image Time Series with Transformers
• Deep Learning for MRI reconstruction: about hallucinations and instabilities
• Self-supervised Learning of Visual Representations

The tutorials are based on recent publications in top-tier scientific journals or conferences. They reproduce parts or variants of their numerical experiments.

Guides

• Processing multispectral satellite images with Python
• 2D Discrete Fourier Transform

Evaluation

In groups of two, you will be randomly assigned one of the tutorial. You will have to complete the tutorial (in case you had not finished it during class), and reproduce another experiment of your choice from the original paper. You will be evaluated on the quality of the code, the presentation of your experiments, and their interpretation.