Long-tailed visual recognition with deep models: A methodological survey and evaluation

Fu, Yu; Xiang, Liuyu; Zahid, Yumna; Ding, Guiguang; Mei, Tao; Shen, Qiang; Han, Jungong

doi:10.1016/j.neucom.2022.08.031

Cited by 20 publications

(13 citation statements)

References 110 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…As early as 1987, research by Biederman [1] found that humans can identify an average of 10,000 to 30,000 things in our lifetime. First, the number of samples in the real world is consistent with a long-tail distribution [2], as shown in Figure 1, where only a few categories have enough samples, and the vast majority have tiny sample sizes. Moreover, humans do not need hundreds or thousands of data when learning a new concept.…”

Section: Introduction 1research Background and Significancementioning

confidence: 53%

A Survey of Few-Shot Learning Research Based on Deep Neural Network

Wu¹

2022

FCIS

View full text Add to dashboard Cite

With the successful development of deep learning techniques in recent years, deep neural networks have achieved excellent results in both computer vision and natural language processing by relying on large-scale datasets but still face significant challenges in solving the problem of learning from few-shot. Inspired by the ability of humans to learn to recognize objects as a way to simulate the cognitive process of learning from a small sample size, few-shot learning is a hot topic of research in deep neural networks today. It is also a significant and challenging problem. This paper first introduces the research background and definition of few-shot learning, introduces the relevant models, and summarizes and analyzes the common approaches to the problem of few-shot learning based on deep neural networks at the present stage, which are divided into four types: data augmentation, model fine-tuning, metric learning and meta-learning. Finally, popular datasets for few-shot learning are described, the paper is concluded and future research directions are discussed.

show abstract

Section: Introduction 1research Background and Significancementioning

confidence: 53%

A Survey of Few-Shot Learning Research Based on Deep Neural Network

Wu¹

2022

FCIS

View full text Add to dashboard Cite

show abstract

“…Long-tailed datasets introduce significant bias in the recognition of tail classes in deep learning models [1,23], which attracts widespread attention and research interest from scholars. The current solutions primarily focus on addressing the impact of class imbalance and can be broadly categorized into the following directions:…”

Section: Long-tailed Visual Recognitionmentioning

confidence: 99%

“…The distribution of different classes of geographical objects often exhibits a characteristic of imbalance, naturally leading to the manifestation of long-tailed distributions in many datasets [1,2]. In these datasets, head classes typically have a large number of samples, while the tail classes are characterized by a comparatively lower sample count [3].…”

Section: Introductionmentioning

confidence: 99%

Long-Tailed Effect Study in Remote Sensing Semantic Segmentation Based on Graph Kernel Principles

Cui,

Feng,

Chen

et al. 2024

Remote Sensing

View full text Add to dashboard Cite

The performance of semantic segmentation in remote sensing, based on deep learning models, depends on the training data. A commonly encountered issue is the imbalanced long-tailed distribution of data, where the head classes contain the majority of samples while the tail classes have fewer samples. When training with long-tailed data, the head classes dominate the training process, resulting in poorer performance in the tail classes. To address this issue, various strategies have been proposed, such as resampling, reweighting, and transfer learning. However, common resampling methods suffer from overfitting to the tail classes while underfitting the head classes, and reweighting methods are limited in the extreme imbalanced case. Additionally, transfer learning tends to transfer patterns learned from the head classes to the tail classes without rigorously validating its generalizability. These methods often lack additional information to assist in the recognition of tail class objects, thus limiting performance improvements and constraining generalization ability. To tackle the abovementioned issues, a graph neural network based on the graph kernel principle is proposed for the first time. By leveraging the graph kernel, structural information for tail class objects is obtained, serving as additional contextual information beyond basic visual features. This method partially compensates for the imbalance between tail and head class object information without compromising the recognition accuracy of head classes objects. The experimental results demonstrate that this study effectively addresses the poor recognition performance of small and rare targets, partially alleviates the issue of spectral confusion, and enhances the model’s generalization ability.

show abstract

“…In the training phase, a hybrid training strategy is proposed to boost the assessment performance: the pre-trained model is first fine-tuned on all 14 types of actions in an unsupervised manner [7], followed by fine-tuning on each type of action separately. To re-balance imbalance samples, we adopt a re-sampling strategy [12] to train the network.…”

Section: Implementation Detailsmentioning

confidence: 99%