RGTransformer: Region-Graph Transformer for Image Representation and Few-Shot Classification

AMPL: An adaptive meta-prompt learner for few-shot image classification.

Recent papers have suggested that transfer learning can outperform sophisticated meta-learning methods for few-shot image classification. We take this hypothesis to its logical conclusion, and suggest the use of an ensemble of high-quality, pre-trained feature extractors for few-shot image classification. We show experimentally that a library of pre-trained feature extractors combined with a simple feed-forward network learned with an L2-regularizer can be an excellent option for solving cross-domain few-shot image classification. Our experimental results suggest that this simple approach far outperforms several well-established meta-learning algorithms.

Few-shot image classification via hybrid representation

Revisiting metric learning for few-shot image classification

KLSANet: Key local semantic alignment Network for few-shot image classification

Metric learning based few-shot image classification has recently received much attention for simplicity and efficiency. Its performance depends highly on the feature extractor and classifier. In this paper, we propose an alternative metric learning method for few-shot image classification––Res-SVDNet. Taking the advantages of good representation and anti-overfitting, the proposed Res-SVDNet utilizes ResNet-18 architecture as the backbone to extract image features. As for the classifier, Euclidean distance is employed to meet the few-shot classification preference for simple inductive bias. Considering the requirement of Euclidean distance for orthogonality, singular value decomposition (SVD) is further involved between feature extraction and classification to improve the classification performance of Res-SVDNet. Experimental results on Mini-ImageNet and Tired-ImageNet demonstrate that the proposed Res-SVDNet outperforms the state-of-the-art methods such as prototypical networks by 2%~6%.

TST_MFL: Two-stage training based metric fusion learning for few-shot image classification

Humans can obtain the knowledge of novel visual concepts from language descriptions, and we thus use the few-shot image classification task to investigate whether a machine learning model can have this capability. Our proposed model, LIDE (Learning from Image and DEscription), has a text decoder to generate the descriptions and a text encoder to obtain the text representations of machine- or user-generated descriptions. We confirmed that LIDE with machine-generated descriptions outperformed baseline models. Moreover, the performance was improved further with high-quality user-generated descriptions. The generated descriptions can be viewed as the explanations of the model's predictions, and we observed that such explanations were consistent with prediction results. We also investigated why the language description improved the few-shot image classification performance by comparing the image representations and the text representations in the feature spaces.

Few-shot image classification commits to recognizing new concepts from limited annotated samples. Our insight is to obtain a sufficiently powerful embedding network (PEN) to solve few-shot classification tasks. We propose a method to tackle the few-shot classification tasks, namely PENs for few-shot image classification. The key core of PEN is gaining a well-trained embedding network that is capable of extracting strong discriminating representations to represent an image by utilizing two strategies. One strategy is that the multi-scale feature maps are fused instead of only utilizing the final top-level feature maps. We consider that low-level features also play an important role instead of only utilizing top-level representations. Another significant strategy is knowledge distillation (KD). The characteristics of KD can help us get better performance of an embedding network to extract features. Finally, a distance function is employed to classify unlabeled samples. Comprehensive experiments are conducted on few-shot benchmarks. Our method achieves promising performances. The results demonstrate that KD and future fusion are beneficial to gain an expected embedding network for few-shot classification tasks.

Few-shot image classification aims to recognize classes with limited labeled data. Many works have been proposed to solve this problem. Recently, the feature map reconstruction network has attracted great attention. However, due to the scarcity of data, it is difficult to find an exact solution to the reconstructed model. We try to use statistical information to draw the support set distribution and query set distribution closer, which helps the model obtain a precise solution. We consider first-order information, calculate the mean of each set and eliminate the bias between them. In addition to one-dimensional information, high-dimensional information (i.e. the covariance matrix) is used to represent the data distribution and make Euclidean distance of them as the loss. Our statistical information feature reconstruction network(SI-FRN) does not bring in any other supervised information. Compared to recent state-of-the-art few-shot image classification methods with reconstruction, our SI-FRN gets excellent performance on four benchmarks: mini-ImageNet, tiered-ImageNet, CUB and Aircraft.

SCL: Self-supervised contrastive learning for few-shot image classification

The problem of few-shot image classification has become a popular research area in the field of imaging. In order to improve the accuracy of few-shot image classification, many learning methods have been proposed, among which there are four main approaches: meta-based learning, data augmentation-based, migration-based learning and metric-based learning. In this paper, we propose a modified earth mover’s distance (MEMD) based on the metric learning approach, which has received much attention due to its simple structure and accurate classification. Similarly, MEMD constructs correlations between image regions, using such correlations to characterise the class of the image. MEMD generates a stream of best matches between image regions, and this stream of matches represents the similarity of the classified images. the MEMD algorithm requires the generation of feature weights for image regions, and EMD uses a mechanism of cross-referenced citations to generate uniform consultation weights. in contrast to EMD, MEMD generates inconsistent reference weights based on the similarity of regions. In dealing with the K-Shot problem, we used a learnable class prototype to characterise the class feature vectors. We conducted comprehensive experiments to validate our improved MEMD algorithm and tested it on four popular few-shot datasets. Namely: miniImageNet, tieredImageNet, Fewshot-CIFAR100 (FC100) and the CUB dataset.

Self-supervised learning (SSL) techniques have recently been integrated into the few-shot learning (FSL) framework and have shown promising results in improving the few-shot image classification performance. However, existing SSL approaches used in FSL typically seek the supervision signals from the global embedding of every single image. Therefore, during the episodic training of FSL, these methods cannot capture and fully utilize the local visual information in image samples and the data structure information of the whole episode, which are beneficial to FSL. To this end, we propose to augment the few-shot learning objective with a novel self-supervised Episodic Spatial Pretext Task (ESPT). Specifically, for each few-shot episode, we generate its corresponding transformed episode by applying a random geometric transformation to all the images in it. Based on these, our ESPT objective is defined as maximizing the local spatial relationship consistency between the original episode and the transformed one. With this definition, the ESPT-augmented FSL objective promotes learning more transferable feature representations that capture the local spatial features of different images and their inter-relational structural information in each input episode, thus enabling the model to generalize better to new categories with only a few samples. Extensive experiments indicate that our ESPT method achieves new state-of-the-art performance for few-shot image classification on three mainstay benchmark datasets. The source code will be available at: https://github.com/Whut-YiRong/ESPT.

In recent years, learnable prompts have emerged as a major prompt learning paradigm, enhancing the performance of large-scale vision-language pre-trained models in few-shot image classification. However, enhancing methods are often time-consuming and inflexible because 1) class-specific prompts are inefficient in certain situations; 2) instance-specific prompts are put in a fixed position. To address these issues, inspired by the coarse-to-fine decision-making paradigm of human, we propose an Instance-Aware Hierarchical-Structured Policy (IAHSP) that integrates instance-specific prompt selection and appropriate position selection using a reinforcement learning fashion. Specifically, IAHSP consists of two sub-policies: 1) the root policy selects the most suitable prompt from the prompts pool, and 2) the leaf policy identifies the optimal position for inserting the selected prompt. We train these two policies iteratively with rewards constraining the prompts while maintaining their diversity. Extensive experiments on 11 public benchmarks demonstrate that our IAHSP significantly boosts the few-shot image classification performance of vision-language pre-trained models, while also exhibiting superior generalization performance.

Deep neural networks have achieved remarkable results in large-scale data domain. However, they have not performed well on few-shot image classification tasks. Here we propose a new meta-learning approach composed of an embedding network and a linear classifier learner. During the training phase, our approach (called Transformation Network) learns to learn a classifier by transforming the feature vectors produced by the embedding module. Once trained, a Transformation Network is able to classify images of new classes by the learned classifier. The ability of learning a discriminatively trained classifier could make our architecture adapt fast to new examples from unseen classes. We further describe implementation details upon the architecture convolutional networks and linear transformation operations. We demonstrate that our approach achieves improved performance on few-shot image classification tasks on two benchmarks and a self-made dataset.