AI Paper Atlas

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

Twin Contrastive Learning for Online Clustering

Yunfan Li, Mouxing Yang, Dezhong Peng, Taihao Li, Jiantao Huang, Xi Peng

IJCV 2022 · Label-efficient and robust learning › Recognition and long-tail · approach: Contrastive representation theory · 164 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 self-labeling boosting loss (ablation)TCL · CIFAR-10ACC: 0.865 -> 0.878Improved
add self-supervised contrastive loss to boosting (SL+SCL) (ablation)TCL · CIFAR-10ACC: 0.878 -> 0.887Improved
add the confidence-based boosting stage (ablation)TCL · ImageNet-1KACC: 0.3160 -> 0.3789Improved
add instance-level contrastive head (ablation)TCL · CIFAR-10ACC: 0.658 -> 0.865Improved
add cluster-level contrastive head (ablation)TCL · CIFAR-10ACC: 0.821 -> 0.865Improved
add cluster-level contrastive head (ablation)TCL · StackOverflowACC: 0.732 -> 0.860Improved
add decoupling of instance and cluster heads (ablation)TCL · CIFAR-10ACC: 0.640 -> 0.854Improved
add decoupling of instance and cluster heads (ablation)TCL · ImageNet-DogsACC: 0.368 -> 0.535Improved
use weak+strong instead of weak+weak augmentation (ablation)TCL · CIFAR-10ACC: 0.805 -> 0.865Improved
use weak+strong instead of strong+strong augmentation (ablation)TCL · ImageNet-DogsACC: 0.147 -> 0.549Improved
use mixed weak+strong augmentation for SimCLR+k-means (ablation)SimCLR+k-means · CIFAR-10ACC: 0.782 -> 0.821 (+0.039)Improved
increase ICH dimension from 10 to 64 (ablation)TCL · CIFAR-10ACC: 0.854 -> 0.871Improved

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