Home Research Papers Datasets State of the Art Pricing

Discover, visualize, and connect AI research papers. Explore the latest trends and insights in artificial intelligence research.

Product

Home
Research Papers
About

Support

Contact
Terms of Service
Privacy Policy

© 2026 Papersgraph. All rights reserved.

methodology-11

Novel Class Discovery

3260 papers • 126 benchmarks • 313 datasets

The goal of Novel Class Discovery (NCD) is to identify new classes in unlabeled data, by exploiting prior knowledge from known classes. In this specific setup, the data is split in two sets. The first is a labeled set containing known classes and the second is an unlabeled set containing unknown classes that must be discovered.

(Image credit: Papersgraph)

Benchmarks

These leaderboards are used to track progress in novel-class-discovery-33

Trend

Dataset

Best Model

Actions

cifar10

cifar10

cifar100

cifar100

SVHN

SVHN

Libraries

i

Use these libraries to find novel-class-discovery-33 models and implementations

Datasets

CIFAR-10

CIFAR-100

SVHN

Subtasks

No subtasks available.

Most implemented papers

Parametric Classification for Generalized Category Discovery: A Baseline Study

Xiaojuan Qi, Bingchen Zhao, Xin Wen•Sun Nov 20 2022

This study investigates the failure of parametric classifiers, verifies the effectiveness of previous design choices when high-quality supervision is available, and identifies unreliable pseudo-labels as a key problem.

128

Content

Introduction Benchmarks Datasets Subtasks Libraries Papers

0

Class-relation Knowledge Distillation for Novel Class Discovery

Xuming He, Peiyan Gu, Chuyu Zhang, Rui Xu•Mon Jul 17 2023

A novel knowledge distillation framework is proposed, which utilizes the authors' class-relation representation to regularize the learning of novel classes and develops a learnable weighting function for the regularization, which adaptively promotes knowledge transfer based on the semantic similarity between the novel and known classes.

33 0

Encoding Structure-Texture Relation with P-Net for Anomaly Detection in Retinal Images

Jiang Liu, Shenghua Gao, Jianlong Yang, Wen Liu, Weixin Luo, Kang Zhou, Yuting Xiao, Jun Cheng, Zaiwang Gu•Sat Aug 08 2020

This work first extracts the structure of the retinal images, then it combines both the structure features and the last layer features extracted from original health image to reconstruct the original input healthy image, and measures the difference between structure extracted from Original and the reconstructed image.

133 0

Meta Discovery: Learning to Discover Novel Classes given Very Limited Data

Tongliang Liu, Bo Han, Gang Niu, Feng Liu, Haoang Chi, Wenjing Yang, L. Lan•Sun Feb 07 2021

This paper demystify assumptions behind NCD and finds that high-level semantic features should be shared among the seen and unseen classes, which proves that NCD is theoretically solvable under certain assumptions and can be naturally linked to meta-learning that has exactly the same assumption as NCD.

51 0

Neighborhood Contrastive Learning for Novel Class Discovery

N. Sebe, E. Ricci, Zhun Zhong, Enrico Fini, Subhankar Roy, Zhiming Luo•Mon May 31 2021

This paper addresses Novel Class Discovery (NCD), the task of unveiling new classes in a set of unlabeled samples given a labeled dataset with known classes, and builds a new framework, named Neighborhood Contrastive Learning (NCL), to learn discriminative representations that are important to clustering performance.

180 0

AutoNovel: Automatically Discovering and Learning Novel Visual Categories

Andrew Zisserman, Sylvestre-Alvise Rebuffi, A. Vedaldi, Sébastien Ehrhardt, Kai Han•Wed Jun 23 2021

This work suggests that the common approach of bootstrapping an image representation using the labelled data only introduces an unwanted bias, and that this can be avoided by using self-supervised learning to train the representation from scratch on the union of labelled and unlabelled data.

182 0

A Unified Objective for Novel Class Discovery

Stéphane Lathuilière, E. Ricci, E. Sangineto, Zhun Zhong, Enrico Fini, Moin Nabi•Wed Aug 18 2021

A UNified Objective function (UNO) for discovering novel classes, with the explicit purpose of favoring synergy between supervised and unsupervised learning, and outperforms the state of the art on several benchmarks.

233 0

Novel Class Discovery in Semantic Segmentation

N. Sebe, Yuyang Zhao, Zhun Zhong, G. Lee•Thu Dec 02 2021

This work introduces a new setting of Novel Class Discovery in Semantic Segmentation (NCDSS), which aims at segmenting unlabeled images containing new classes given prior knowledge from a labeled set of disjoint classes, and proposes the Entropy-based Uncertainty Modeling and Self-training (EUMS) framework to overcome noisy pseudo-labels.

38 0

Spacing Loss for Discovering Novel Categories

K. J. Joseph, V. Balasubramanian, Kai Han, Soma Biswas, S. Paul, G. Aggarwal, Piyush Rai•Thu Apr 21 2022

This work characterize existing NCD approaches into singlestage and two-stage methods based on whether they require access to labeled and unlabeled data together while discovering new classes and devise a simple yet powerful loss function that enforces separability in the latent space using cues from multi-dimensional scaling, which is referred to as Spacing Loss.

18 0

Adding a benchmark result helps the community track progress.

Novel Class Discovery | State-of-the-Art