AI Paper Atlas

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

A Unified Objective for Novel Class Discovery

Enrico Fini, Enver Sangineto, Stéphane Lathuilière, Zhun Zhong, Moin Nabi, Elisa Ricci

ICCV 2021 · Label-efficient and robust learning › Semi-supervised learning · approach: Confidence-based pseudo-label selection · 210 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 logit concatenation into one shared softmax (unified objective) (ablation)UNO · CIFAR10 task-agnostic all classesaccuracy: 65.9 -> 93.4Improved
add logit concatenation into one shared softmax (unified objective) (ablation)UNO · CIFAR100-50 task-agnostic all classesaccuracy: 54.3 -> 61.1Improved
add overclustering heads (ablation)UNO · CIFAR10 task-agnostic all classesaccuracy: 92.0 -> 93.4Improved
add overclustering heads (ablation)UNO · CIFAR100-50 task-agnostic all classesaccuracy: 59.0 -> 61.1Improved
add overclustering heads (ablation)UNO · CIFAR100-50 task-aware labeled classesaccuracy: 78.9 -> 78.8No clear effect
use strong instead of weak augmentation (ablation)UNO · CIFAR10 task-agnostic all classesaccuracy: 91.8 -> 93.4Improved
use strong instead of weak augmentation (ablation)UNO · CIFAR100-50 task-agnostic all classesaccuracy: 59.1 -> 61.1Improved
apply strong augmentations to RSRS · novel class discoveryclustering accuracyNo clear effect
swapped prediction instead of averaging logits (ablation)UNO · CIFAR10 (train)clustering accuracy: 94.4 -> 96.1Improved
swapped prediction instead of averaging logits (ablation)UNO · CIFAR100-50 (train)clustering accuracy: 52.2 -> 52.9Improved
swapped prediction instead of averaging pseudo-labels (ablation)UNO · CIFAR100-50 (train)clustering accuracy: 52.5 -> 52.9No clear effect
self-supervised pretraining instead of supervised pretraining (ablation)UNO · novel class discoveryclustering accuracyNo clear effect

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