AI Paper Atlas

Research map / Label-efficient and robust learning / Recognition and long-tail

CDUL: CLIP-Driven Unsupervised Learning for Multi-Label Image Classification

Rabab Abdelfattah, Qing Guo, Xiaoguang Li, Xiaofeng Wang, Song Wang

ICCV 2023 · Label-efficient and robust learning › Recognition and long-tail · approach: Partial and multi-label learning · 71 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 local snippet alignment (3x3 crops) to CLIP global alignment for pseudo labels (ablation)CLIP pseudo labels · VOC 2012 trainmAP: 85.3 -> 90.3Improved
add local snippet alignment to CLIP global alignment for pseudo labels (ablation)CLIP pseudo labels · COCO trainmAP: 65.4 -> 72.8Improved
use min-max aggregator instead of average aggregator (ablation)global-local pseudo labels · VOC 2012 trainmAP: 88.5 -> 90.3Improved
use min-max aggregator instead of max aggregator (ablation)global-local pseudo labels · COCO trainmAP: 71.6 -> 72.8Improved
use larger CLIP backbone (ResNet-50x64) for pseudo labels (ablation)CDUL · PASCAL VOC 2007mAP: 86.9 -> 89.0Improved
add Gaussian weighting module in gradient-alignment training (ablation)CDUL · VOC 2012 and COCOmAP: -0.5% VOC 2012, -0.4% COCO when removedImproved
use soft pseudo labels instead of hard pseudo labels (ablation)CDUL · VOC 2012 and COCOmAP: -0.9% VOC 2012, -1.2% COCO with hard labelsImproved
train a ResNet-50 student on CLIP pseudo labels instead of using CLIP-GLA directly at inference (ablation)CDUL · COCOmAP: 63.6 -> 67.1Improved
train classifier with gradient-alignment on CLIP pseudo labels vs raw CLIPCDUL · MS-COCOmAP: 65.3 -> 69.2Improved

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