Abstract:Deep long-tailed learning, one of the most challenging problems in visual recognition, aims to train well-performing deep models from a large number of images that follow a long-tailed class distribution. In the last decade, deep learning has emerged as a powerful recognition model for learning high-quality image representations and has led to remarkable breakthroughs in generic visual recognition. However, long-tailed class imbalance, a common problem in practical visual recognition tasks, often limits the practicality of deep network based recognition models in real-world applications, since they can be easily biased towards dominant classes and perform poorly on tail classes. To address this problem, a large number of studies have been conducted in recent years, making promising progress in the field of deep long-tailed learning. Considering the rapid evolution of this field, this paper aims to provide a comprehensive survey on recent advances in deep long-tailed learning. To be specific, we group existing deep long-tailed learning studies into three main categories (i.e., class re-balancing, information augmentation and module improvement), and review these methods following this taxonomy in detail. Afterward, we empirically analyze several state-of-the-art methods by evaluating to what extent they address the issue of class imbalance via a newly proposed evaluation metric, i.e., relative accuracy. We conclude the survey by highlighting important applications of deep long-tailed learning and identifying several promising directions for future research.

Deep Active Learning over the Long Tail

Effective Active Learning Method for Spiking Neural Networks.

Deep Individual Active Learning: Safeguarding Against Out-of-Distribution Challenges in Neural Networks

Deep Active Learning with Contrastive Learning Under Realistic Data Pool Assumptions

Deep Bayesian Active Learning with Image Data

Deep Active Learning by Leveraging Training Dynamics

Nearest Neighbor Classifier Embedded Network for Active Learning

Deep Active Learning with a Neural Architecture Search

Adversarial Active Learning for Deep Networks: a Margin Based Approach

Deep Active Learning: A Reality Check

Deep Adaptive Temporal Pooling for Activity Recognition

Does Deep Active Learning Work in the Wild?

Transductive Active Learning: Theory and Applications

A Deep Learning Model for Long-Tail Visual Recognition

Diminishing Uncertainty Within the Training Pool: Active Learning for Medical Image Segmentation

Streaming Active Learning with Deep Neural Networks

Amortized Active Learning for Nonparametric Functions

Cost-Effective Active Learning for Deep Image Classification

Dual Adversarial Network for Deep Active Learning

AnchorAL: Computationally Efficient Active Learning for Large and Imbalanced Datasets

Deep Long-Tailed Learning: A Survey