AI Paper Atlas

Research map / Label-efficient and robust learning / Semi-supervised learning

Harvard Glaucoma Detection and Progression: A Multimodal Multitask Dataset and Generalization-Reinforced Semi-Supervised Learning

Yan Luo, Min Shi, Yu Tian, Tobias Elze, Mengyu Wang

ICCV 2023 · Label-efficient and robust learning › Semi-supervised learning · approach: GAN-based semi-supervised learning · 38 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add weak augmentations to Pseudo-Sup (ablation)Pseudo-Sup · Harvard-GDP glaucoma detectionAUC: 0.8727 -> 0.8818Improved
add weak augmentations to Pseudo-Sup on progression forecasting (ablation)Pseudo-Sup · Harvard-GDP progression forecasting, MD FastAUC: 0.7004 -> 0.7215Improved
use Pseudo-Sup instead of supervised baseline on glaucoma detectionsupervised EfficientNetV2-S · Harvard-GDP glaucoma detectionaccuracy: 0.8131 -> 0.8124No clear effect
use semi-supervised MixMatch/ReMixMatch/DASO instead of supervised baseline on glaucoma detectionsupervised EfficientNetV2-S · Harvard-GDP glaucoma detectionaccuracy: 0.8131 -> 0.7444 (DASO)Hurt
use Pseudo-Sup instead of supervised baseline on MD Fast progressionsupervised EfficientNetV2-S · Harvard-GDP progression forecasting, MD FastAUC: 0.6677 -> 0.7004Improved
use Pseudo-Sup with multimodal RNFLT+VF inputPseudo-Sup · Harvard-GDP progression forecasting, MD FastAUC: 0.7004 -> 0.7265Improved
Pseudo-Sup on LAG fundus dataset vs FixMatchFixMatch · LAG fundus glaucoma detectionAUC: 0.9198 -> 0.9858Improved
use transformer backbones instead of CNNs for supervised benchmarksupervised glaucoma detection · Harvard-GDP released 1,000 patientsAUC: 0.90 (EfficientNet) -> 0.75 (ViT)Hurt
set policy time window beta to 50 and discount rate gamma to 0.9 (ablation)Pseudo-Sup · Harvard-GDP TD progressionAUCImproved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy