Semi-Supervised Image Classification

Edit

65 papers with code • 34 benchmarks • 9 datasets

Semi-supervised image classification leverages unlabelled data as well as labelled data to increase classification performance.

You may want to read some blog posts to get an overview before reading the papers and checking the leaderboards:

An overview of proxy-label approaches for semi-supervised learning - Sebastian Ruder
Semi-Supervised Learning in Computer Vision - Amit Chaudhary

( Image credit: Self-Supervised Semi-Supervised Learning )

Benchmarks

Add a Result

Dataset	Best Model	Paper Title	Compare
ImageNet - 10% labeled data	SimCLRv2 self-distilled (ResNet-152 x3, SK)	Big Self-Supervised Models are Strong Semi-Supervised Learners	See all
ImageNet - 1% labeled data	SimCLRv2 self-distilled (ResNet-152 x3, SK)	Big Self-Supervised Models are Strong Semi-Supervised Learners	See all
CIFAR-10, 4000 Labels	LaplaceNet (WRN-28-8)	LaplaceNet: A Hybrid Energy-Neural Model for Deep Semi-Supervised Classification	See all
SVHN, 1000 labels	Meta Pseudo Labels	Meta Pseudo Labels	See all
SVHN, 250 Labels	EnAET	EnAET: A Self-Trained framework for Semi-Supervised and Supervised Learning with Ensemble Transformations	See all
cifar-100, 10000 Labels	LaplaceNet (WRN-28-8)	LaplaceNet: A Hybrid Energy-Neural Model for Deep Semi-Supervised Classification	See all
CIFAR-10, 250 Labels	SelfMatch	SelfMatch: Combining Contrastive Self-Supervision and Consistency for Semi-Supervised Learning	See all
CIFAR-10, 1000 Labels	MixMatch	MixMatch: A Holistic Approach to Semi-Supervised Learning	See all
STL-10, 1000 Labels	FixMatch (CTA)	FixMatch: Simplifying Semi-Supervised Learning with Consistency and Confidence	See all
CIFAR-10, 40 Labels	SelfMatch	SelfMatch: Combining Contrastive Self-Supervision and Consistency for Semi-Supervised Learning	See all
SVHN, 500 Labels	Triple-GAN-V2 (CNN-13)	Triple Generative Adversarial Networks	See all
cifar10, 250 Labels	ReMixMatch	ReMixMatch: Semi-Supervised Learning with Distribution Alignment and Augmentation Anchoring	See all
CIFAR-10, 2000 Labels	MixMatch	MixMatch: A Holistic Approach to Semi-Supervised Learning	See all
Mini-ImageNet, 4000 Labels	SimPLE	SimPLE: Similar Pseudo Label Exploitation for Semi-Supervised Classification	See all
STL-10	EnAET	EnAET: A Self-Trained framework for Semi-Supervised and Supervised Learning with Ensemble Transformations	See all
Mini-ImageNet, 10000 Labels	FeatMatch	FeatMatch: Feature-Based Augmentation for Semi-Supervised Learning	See all
Mini-ImageNet, 1000 Labels	SemCo (μ=3)	All Labels Are Not Created Equal: Enhancing Semi-supervision via Label Grouping and Co-training	See all
CIFAR-100, 1000 Labels	EnAET	EnAET: A Self-Trained framework for Semi-Supervised and Supervised Learning with Ensemble Transformations	See all
CIFAR-100, 5000Labels	EnAET	EnAET: A Self-Trained framework for Semi-Supervised and Supervised Learning with Ensemble Transformations	See all
CIFAR-100, 400 Labels	FixMatch (CTA)	FixMatch: Simplifying Semi-Supervised Learning with Consistency and Confidence	See all
CIFAR-100, 2500 Labels	FixMatch (CTA)	FixMatch: Simplifying Semi-Supervised Learning with Consistency and Confidence	See all
SVHN, 40 Labels	FixMatch (CTA)	FixMatch: Simplifying Semi-Supervised Learning with Consistency and Confidence	See all
STL-10, 5000 Labels	MixMatch	MixMatch: A Holistic Approach to Semi-Supervised Learning	See all
SVHN, 2000 Labels	MixMatch	MixMatch: A Holistic Approach to Semi-Supervised Learning	See all
SVHN, 4000 Labels	MixMatch	MixMatch: A Holistic Approach to Semi-Supervised Learning	See all
CIFAR-10, 500 Labels	MixMatch	MixMatch: A Holistic Approach to Semi-Supervised Learning	See all
Caltech-101	UL-Hopfield (ULH)	Unsupervised Learning using Pretrained CNN and Associative Memory Bank	See all
Caltech-101, 202 Labels	UL-Hopfield (ULH)	Unsupervised Learning using Pretrained CNN and Associative Memory Bank	See all
Caltech-256, 1024 Labels	UL-Hopfield (ULH)	Unsupervised Learning using Pretrained CNN and Associative Memory Bank	See all
Caltech-256	UL-Hopfield (ULH)	Unsupervised Learning using Pretrained CNN and Associative Memory Bank	See all
CIFAR-10, 20 Labels	SimCLR (CoMatch)	CoMatch: Semi-supervised Learning with Contrastive Graph Regularization	See all
CIFAR-10, 80 Labels	SimCLR (CoMatch)	CoMatch: Semi-supervised Learning with Contrastive Graph Regularization	See all
cifar-10, 10 Labels	BOSS	Building One-Shot Semi-supervised (BOSS) Learning up to Fully Supervised Performance	See all
CIFAR-100, 4000 Labels	LaplaceNet (WRN-28-8)	LaplaceNet: A Hybrid Energy-Neural Model for Deep Semi-Supervised Classification	See all

Show all 34 benchmarks

Collapse benchmarks

Datasets

Greatest papers with code

Improved Techniques for Training GANs

tensorflow/models • • NeurIPS 2016

We present a variety of new architectural features and training procedures that we apply to the generative adversarial networks (GANs) framework.

Ranked #11 on Conditional Image Generation on CIFAR-10

Conditional Image Generation Semi-Supervised Image Classification

61,324

Paper
Code

Milking CowMask for Semi-Supervised Image Classification

google-research/google-research • • 26 Mar 2020

Using it to provide perturbations for semi-supervised consistency regularization, we achieve a state-of-the-art result on ImageNet with 10% labeled data, with a top-5 error of 8. 76% and top-1 error of 26. 06%.

Ranked #9 on Semi-Supervised Image Classification on ImageNet - 10% labeled data

General Classification Semi-Supervised Image Classification

18,537

Paper
Code

Meta Pseudo Labels

google-research/google-research • • CVPR 2021

We present Meta Pseudo Labels, a semi-supervised learning method that achieves a new state-of-the-art top-1 accuracy of 90. 2% on ImageNet, which is 1. 6% better than the existing state-of-the-art.

Ranked #1 on Semi-Supervised Image Classification on SVHN, 1000 labels

Meta-Learning Semi-Supervised Image Classification

18,536

Paper
Code

mixup: Beyond Empirical Risk Minimization

rwightman/pytorch-image-models • • ICLR 2018

We also find that mixup reduces the memorization of corrupt labels, increases the robustness to adversarial examples, and stabilizes the training of generative adversarial networks.

Ranked #7 on Domain Generalization on ImageNet-A

Domain Generalization Semi-Supervised Image Classification

11,791

Paper
Code

Semi-Supervised Learning with Context-Conditional Generative Adversarial Networks

eriklindernoren/PyTorch-GAN • • 19 Nov 2016

We introduce a simple semi-supervised learning approach for images based on in-painting using an adversarial loss.

Ranked #5 on Semi-Supervised Image Classification on STL-10, 1000 Labels

Semi-Supervised Image Classification

9,806

Paper
Code

Bootstrap your own latent: A new approach to self-supervised Learning

deepmind/deepmind-research • • 13 Jun 2020

From an augmented view of an image, we train the online network to predict the target network representation of the same image under a different augmented view.

Ranked #5 on Semi-Supervised Image Classification on ImageNet - 1% labeled data

Representation Learning Self-Supervised Image Classification +2

7,218

Paper
Code

Big Self-Supervised Models are Strong Semi-Supervised Learners

google-research/simclr • • NeurIPS 2020

The proposed semi-supervised learning algorithm can be summarized in three steps: unsupervised pretraining of a big ResNet model using SimCLRv2, supervised fine-tuning on a few labeled examples, and distillation with unlabeled examples for refining and transferring the task-specific knowledge.

Ranked #1 on Semi-Supervised Image Classification on ImageNet - 10% labeled data

Self-Supervised Image Classification Semi-Supervised Image Classification

2,368

Paper
Code

A Simple Framework for Contrastive Learning of Visual Representations

google-research/simclr • • ICML 2020

This paper presents SimCLR: a simple framework for contrastive learning of visual representations.

Ranked #5 on Semi-Supervised Image Classification on ImageNet - 10% labeled data

Contrastive Learning Self-Supervised Image Classification +2

2,368

Paper
Code

Improved Regularization of Convolutional Neural Networks with Cutout

PaddlePaddle/PaddleClas • • 15 Aug 2017

Convolutional neural networks are capable of learning powerful representational spaces, which are necessary for tackling complex learning tasks.

Ranked #3 on Semi-Supervised Image Classification on STL-10

Domain Generalization Image Augmentation +1

2,131

Paper
Code

Barlow Twins: Self-Supervised Learning via Redundancy Reduction

facebookresearch/vissl • • 4 Mar 2021

This causes the embedding vectors of distorted versions of a sample to be similar, while minimizing the redundancy between the components of these vectors.

Ranked #1 on Image Classification on Places205

General Classification Object Detection +3

1,810

Paper
Code

Semi-Supervised Image Classification Edit

Benchmarks Add a Result

Datasets

Greatest papers with code

Semi-Supervised Image Classification

Edit

Benchmarks

Add a Result