AI Paper Atlas

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

SCOMatch: Alleviating Overtrusting in Open-set Semi-supervised Learning

Zerun Wang, Liuyu Xiang, Lang Huang, Jiafeng Mao, Ling Xiao, Toshihiko Yamasaki

ECCV 2024 · Label-efficient and robust learning › Semi-supervised learning · approach: Safe and realistic semi-supervised learning · 6 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
use a single (K+1)-class head instead of two separate heads (ablation)SCOMatch · CIFAR-10 (6 ID classes, 25 labels)open-set accuracy: 54.3 -> 95.4Improved
use a single head instead of two separate heads (ablation)SCOMatch · CIFAR-100 (55 ID classes, 25 labels)AUC: 72.1 -> 90.6Improved
add close-set self-training on first K dimensions (ablation)SCOMatch · CIFAR-100 (55 ID classes, 25 labels)close-set accuracy: 68.7 -> 73.3Improved
add close-set self-training (ablation)SCOMatch · CIFAR-10 (6 ID classes, 25 labels)close-set accuracy: 93.2 -> 94.4Improved
add open-set self-training (ablation)SCOMatch · CIFAR-10 (6 ID classes, 25 labels)open-set accuracy: 80.6 -> 95.4Improved
add open-set self-training (ablation)SCOMatch · CIFAR-100 (55 ID classes, 25 labels)AUC: 75.0 -> 90.6Improved
add OOD memory queue with (K+1)-class SSL over FixMatch (ablation)FixMatch · CIFAR-10 (6 ID classes, 25 labels)AUC: 37.1 -> 99.3Improved
SCOMatch versus strongest prior OSSL methodMa et al. · TinyImageNet (mismatch 0.3)close-set accuracy: 40.8 -> 54.2Improved
SCOMatch versus OpenMatch on CIFAR-100 80/20 with 50 labelsOpenMatch · CIFAR-100 (80/20 classes, 50 labels)AUC: 86.2 -> 85.4Hurt
treat OOD as a single extra class rather than multiple categoriesSCOMatch · OSSL benchmarksclose-set accuracyImproved

Later papers that compared with this method

2 later accepted papers report SCOMatch as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 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