AI Paper Atlas

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

Combating Noise: Semi-supervised Learning by Region Uncertainty Quantification

Zhenyu Wang, Ya-Li Li, Ye Guo, Shengjin Wang

NeurIPS 2021 · Label-efficient and robust learning › Semi-supervised learning · approach: Confidence-based pseudo-label selection · 9 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 region uncertainty quantification to construct soft targets and train with multi-peak class probabilitiesSemi-supervised Faster R-CNN · PASCAL VOC 2007 test, VOC07 labeled and VOC12 unlabeledAP50:95: 45.2 -> 49.3Improved
Use region uncertainty quantification to construct soft targets and train with multi-peak class probabilitiesSemi-supervised Faster R-CNN · COCO minival, coco-35 labeled and coco-80 unlabeledAP50:95: 33.1 -> 35.5Improved
Add cross-category uncertainty (CCU) (ablation)Semi-supervised detector · VOC07 testAP50:95: 45.2 -> 46.2Improved
Add cross-category uncertainty (CCU) (ablation)Fully supervised detector · VOC07 testAP50:95: 43.1 -> 42.5Hurt
Add uncertainty-aware soft target (UST) (ablation)Semi-supervised detector · VOC07 testAP50:95: 45.2 -> 48.2Improved
Add both CCU and UST (ablation)Semi-supervised detector · VOC07 testAP50:95: 45.2 -> 49.3Improved
Add cross-category uncertainty (CCU) (ablation)Semi-supervised detector · COCO minivalAP50:95: 33.1 -> 34.3Improved
Add cross-category uncertainty (CCU) (ablation)Fully supervised detector · COCO minivalAP50:95: 31.3 -> 31.2No clear effect
Add uncertainty-aware soft target (UST) (ablation)Semi-supervised detector · COCO minivalAP50:95: 33.1 -> 33.7Improved
Add both CCU and UST (ablation)Semi-supervised detector · COCO minivalAP50:95: 33.1 -> 35.5Improved
Use a 3x rather than 1x training scheduleOurs on COCO minival with coco-115/coco-120 · COCO minivalAP50:95: 40.6 -> 41.7Improved
Add strong data augmentation to the 3x training scheduleOurs on COCO minival with coco-115/coco-120 · COCO minivalAP50:95: 41.7 -> 43.2Improved

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