Project
Classification IRM semi-supervisée
Prototype de computer vision pour classifier des IRM cérébrales avec peu d'annotations expertes, en combinant embeddings ResNet, clustering et entraînement semi-supervisé.
PythonComputer VisionPyTorchResNetSemi-Supervised LearningClustering
Role
Computer vision pipeline design, MRI dataset validation, feature extraction, clustering analysis, and semi-supervised CNN training.
Outcome
A research prototype comparing supervised and weak-label-assisted ResNet training for binary brain MRI classification under a constrained annotation budget.