Etudiant : HANA OTHMANE

Filière : Master Big Data et Systèmes intelligents (BDSI)

Encadrant : Pr. TAIRI HAMID

Annèe : 2026

Résumé : Through this project, we are interested in the automated classification of diabetic retinopathy from fundus images based on Deep Learning. To achieve our objective, we carried out a synthesis of some existing image classification and decomposition algorithms to choose the most appropriate and effective ones from the literature. We chose to implement a multi-branch deep learning architecture combining PDE-based image decomposition and a knowledge distillation strategy for diabetic retinopathy classification in fundus images. Training and test results are applied to the APTOS 2019 Dataset.