Self-Supervised Learning for Few-Shot Image Classification

Chen, Da; Chen, Yuefeng; Li, Yuhong; Mao, Feng; He, Yuan; Xue, Hui

doi:10.1109/icassp39728.2021.9413783

Cited by 65 publications

(37 citation statements)

References 18 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…During pre-training phase, we train the encoder on unsupervised Image900-SSL [8] that contains all images from ImageNet1K except miniImageNet. In addition, the classes in Image900-SSL [8] are distinct from the classes present in CUB.…”

Section: Implementation Detailsmentioning

confidence: 99%

“…Researchers have developed algorithms to improve the generalization of few-shot learner. Among them, the meta-learning [3,4,5,6,7,8] and fine-tuning methods [9] achieve excellent performance. Notably, both methods described above use Convolutional Neural Network (CNN) encoders, while Vision Transformer (ViT) [10] generalize better than CNN under multiple distribution shifts, which is demonstrated in previous work [11].…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Improving Few-Shot Learning with Vision Transformer

2022

J. Phys.: Conf. Ser.

View full text Add to dashboard Cite

Vision Transformer (ViT) is emerging as an alternative to convolutional neural network (CNN) for visual recognition and has achieved impressive results, however, due to its data-hungry nature, ViT encoder in few-shot setting remains rarely explored. In this paper, we first propose a simple yet effective baseline that exploits the complementarity of self-supervised learning (SSL) and ViT, enabling the use of standard ViT model as the few-shot learner. Second, based on the baseline, we introduce a novel regularized fine-tuning framework, where the Parametric Instance discrimination (PID) and Base-Novel (BN) regularizations are proposed to reduce the intra-class variance and calibrate the biased distribution of novel classes to further enhance few-shot recognition. We conduct extensive experiments and show that our method can achieve new state-of-the-art performances on two widely used benchmarks.

show abstract

Section: Implementation Detailsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Improving Few-Shot Learning with Vision Transformer

2022

J. Phys.: Conf. Ser.

View full text Add to dashboard Cite

show abstract

“…PT+MAP 17 and LaplacianShot 18 function similarly, however, both propose alternative strategies for distance metrics when considering query and support points. AmdimNet 19 and S2M2 20 , alternatively, leverage self-supervised techniques in order to generate a stronger embedding-space mapping for input data.…”

Section: Transductive and Self-supervised Approaches To Few-shot Lear...mentioning

confidence: 99%

“…Technique Backbone Preprocessing Extra Training Data AmdimNet 19 Self-supervised Metric AmdimNet No Yes EPNet 16 Transductive Metric WRN28-10 No Yes SimpleCNAPS 14 Metric ResNet18 No Yes PT+MAP 17 Metric WRN28-10 Yes No LaplacianShot 18 Metric WRN28-10 No No S2M2R 20 Self-supervised Metric WRN28-10 Yes No Reptile 13 Optimization CONV4 No No MAML 12 Optimization CONV4 No No ProtoNet 15 Metric CONV4 No No Table 1. An overview of the differing details between the models trained and tested.…”

Section: Model Evaluation Table Model Namementioning

confidence: 99%

Automated Human Cell Classification in Sparse Datasets using Few-Shot Learning

Walsh,

Abdelpakey,

Shehata

et al. 2021

Preprint

View full text Add to dashboard Cite

Classifying and analyzing human cells is a lengthy procedure, often involving a trained professional. In an attempt to expedite this process, an active area of research involves automating cell classification through use of deep learning-based techniques. In practice, a large amount of data is required to accurately train these deep learning models. However, due to the sparse human cell datasets currently available, the performance of these models is typically low. This study investigates the feasibility of using few-shot learning-based techniques to mitigate the data requirements for accurate training. The study is comprised of three parts: First, current state-of-the-art few-shot learning techniques are evaluated on human cell classification. The selected techniques are trained on a non-medical dataset and then tested on two out-of-domain, human cell datasets. The results indicate that, overall, the test accuracy of state-of-the-art techniques decreased by at least 30% when transitioning from a non-medical dataset to a medical dataset. Reptile and EPNet were the top performing techniques tested on the BCCD dataset and HEp-2 dataset respectively. Second, this study evaluates the potential benefits, if any, to varying the backbone architecture and training schemes in current state-of-the-art few-shot learning techniques when used in human cell classification. To this end, the best technique identified in the first part of this study, EPNet, is used for experimentation. In particular, the study used 6 different network backbones, 5 data augmentation methodologies, and 2 model training schemes. Even with these additions, the overall test accuracy of EPNet decreased from 88.66% on non-medical datasets to 44.13% at best on the medical datasets. Third, this study presents future directions for using few-shot learning in human cell classification. In general, few-shot learning in its current state performs poorly on human cell classification. The study proves that attempts to modify existing network architectures are not effective and concludes that future research effort should be focused on improving robustness towards out-of-domain testing using optimization-based or self-supervised few-shot learning techniques.

show abstract

“…From this basic learning setting, many extensions have been proposed to improve the performance of metric learning methods. Some of these works focus on pre-training the embedding network [2], others introduce task attention modules [3,11,23], whereas other try to optimize the embeddings [10] and yet others try to use a variety of loss functions [23].…”

Section: Introductionmentioning

confidence: 99%

Finding Significant Features for Few-Shot Learning using Dimensionality Reduction

Mendez-Ruiz,

Gonzalez-Zapata,

Ochoa-Ruiz

et al. 2021

Preprint

View full text Add to dashboard Cite

Few-shot learning is a relatively new technique that specializes in problems where we have little amounts of data. The goal of these methods is to classify categories that have not been seen before with just a handful of samples. Recent approaches, such as metric learning, adopt the meta-learning strategy in which we have episodic tasks conformed by support (training) data and query (test) data. Metric learning methods have demonstrated that simple models can achieve good performance by learning a similarity function to compare the support and the query data. However, the feature space learned by a given metric learning approach may not exploit the information given by a specific few-shot task. In this work, we explore the use of dimension reduction techniques as a way to find task-significant features helping to make better predictions. We measure the performance of the reduced features by assigning a score based on the intra-class and inter-class distance, and selecting a feature reduction method in which instances of different classes are far away and instances of the same class are close. This module helps to improve the accuracy performance by allowing the similarity function, given by the metric learning method, to have more discriminative features for the classification. Our method outperforms the metric learning baselines in the miniImageNet dataset by around 2% in accuracy performance.

show abstract

Self-Supervised Learning for Few-Shot Image Classification

Cited by 65 publications

References 18 publications

Improving Few-Shot Learning with Vision Transformer

Improving Few-Shot Learning with Vision Transformer

Automated Human Cell Classification in Sparse Datasets using Few-Shot Learning

Finding Significant Features for Few-Shot Learning using Dimensionality Reduction

Contact Info

Product

Resources

About