AI Paper Atlas

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

Class Rectification Hard Mining for Imbalanced Deep Learning

Qi Dong, Shaogang Gong, Xiatian Zhu

ICCV 2017 · Label-efficient and robust learning › Recognition and long-tail · approach: Loss design and classifier theory · 206 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 CRL with instance-level hard miningDeepID2 · CelebAaverage class-balanced accuracy: 81 -> 86Improved
use CRL(I) instead of LMLEimbalanced learning on DeepID2 · CelebAaverage class-balanced accuracy: 84 -> 86Improved
use CRL(I) instead of LMLE on the 20 most imbalanced attributesimbalanced learning · CelebAaverage accuracy: 2% -> 7%Improved
use CRL(I) instead of LMLEimbalanced learning · X-Domainaverage class-balanced accuracy: 75.77 -> 79.66Improved
add class-level mining CRL(C) instead of LMLEimbalanced learning · X-Domainaverage class-balanced accuracy: 75.77 -> 78.24Improved
add over-samplingDeepID2 · CelebAaverage class-balanced accuracy: 81 -> 82No clear effect
add down-samplingDeepID2 · CelebAaverage class-balanced accuracy: 81 -> 78Hurt
add down-samplingMTCT · X-Domainaverage class-balanced accuracy: 73.53 -> 26.29Hurt
use instance-level instead of class-level hard mining with relative loss (ablation)CRL · CelebAadditional gain in average accuracy: 4.23 -> 5.85Improved
use instance-level instead of class-level hard mining with relative loss (ablation)CRL · X-Domainadditional gain in average accuracy: 4.71 -> 6.13Improved
use instance-level instead of class-level hard mining with absolute loss (ablation)CRL · CelebAadditional gain in average accuracy: 5.71 -> 5.67No clear effect
use distribution comparison loss instead of relative comparison loss (instance-level) (ablation)CRL · CelebAadditional gain in average accuracy: 5.85 -> 2.12Hurt

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