Semi-Supervised Learning

Pseudo-Label Foreground Quality Weighting for Domain-Generalized Fundus Image Segmentation
Structure-specific pseudo-label quality estimates weight Noisy Student supervision for optic disc and cup segmentation in uncropped fundus images, improving generalization across known and unseen domains, particularly for the optic cup.
Lucas Gabriel Telesco, José Ignacio Orlando
PDF Cite Code Project Project Conference
Pseudo-Label Foreground Quality Weighting for Domain-Generalized Fundus Image Segmentation
Semi-Supervised Multi-Task Learning for Interpretable Quality Assessment of Fundus Images
We propose a semi-supervised multi-task framework for interpretable retinal image quality assessment that uses pseudo-labels for capture conditions, improving overall quality classification on EyeQ and DeepDRiD while reducing the need for detailed expert annotations.
Lucas Gabriel Telesco, Danila Nejamkin, Estefanía Mata, Francisco Filizzola, Kevin Wignall, Lucía Franco Troilo, María De Los Ángeles Cenoz, Melissa Thompson, Mercedes Leguía, Ignacio Larrabide, José Ignacio Orlando
PDF Cite Code Dataset Project DOI arXiv
Semi-Supervised Multi-Task Learning for Interpretable Quality Assessment of Fundus Images