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.

What problem does this paper attempt to address?

### Problems the Paper Attempts to Solve This paper aims to address the key issues in Deep Long-Tailed Learning. In practical visual recognition tasks, training samples often exhibit a long-tailed distribution, where a few classes have a large number of samples, while most classes have only a few samples. This class imbalance problem causes deep network-based recognition models to perform poorly in real-world applications, as these models tend to be biased towards the dominant classes (head classes) and perform poorly on the tail classes. Specifically, the paper focuses on the following aspects: 1. **Class Imbalance Problem**: How to train a well-performing deep model when the class distribution of the training data is extremely imbalanced. 2. **Challenges in Practical Applications**: The long-tailed distribution problem is prevalent in many practical application scenarios, such as face recognition, species classification, medical image diagnosis, and urban scene understanding. Therefore, solving this problem is crucial for improving model performance in these fields. 3. **Review of Existing Methods**: The paper provides a comprehensive review of recent research in the field of deep long-tailed learning, categorizing existing methods into three major types: Class Re-balancing, Information Augmentation, and Module Improvement. 4. **Evaluation and Analysis**: By introducing a new evaluation metric—Relative Accuracy, the paper conducts empirical analysis on several state-of-the-art methods to assess their effectiveness in handling the class imbalance problem. 5. **Future Research Directions**: The paper identifies several important application areas and proposes potential future research directions to advance the development of deep long-tailed learning. ### Summary The goal of this paper is to provide a comprehensive review of deep long-tailed learning, helping researchers better understand and address the challenges posed by long-tailed distributions, thereby improving model performance in practical applications.

Deep Long-Tailed Learning: A Survey

Long-tailed Visual Recognition with Deep Models: A Methodological Survey and Evaluation

A Survey on Long-Tailed Visual Recognition

A Deep Learning Model for Long-Tail Visual Recognition

A Systematic Review on Long-Tailed Learning

A Survey of Deep Long-Tail Classification Advancements

Constructing Balance from Imbalance for Long-Tailed Image Recognition.

Federated deep long-tailed learning: A survey

Bt-Vmf Contrastive and Collaborative Learning for Long-Tailed Visual Recognition

Divide and Retain: A Dual-Phase Modeling for Long-Tailed Visual Recognition

Decoupling Representation and Classifier for Long-Tailed Recognition

BBN: Bilateral-Branch Network with Cumulative Learning for Long-Tailed Visual Recognition

Balanced Contrastive Learning for Long-Tailed Visual Recognition

LTRL: Boosting Long-tail Recognition via Reflective Learning

Self Supervision to Distillation for Long-Tailed Visual Recognition

Investigate the Essence of Long-Tailed Recognition from a Unified Perspective

Revisiting Long-tailed Image Classification: Survey and Benchmarks with New Evaluation Metrics

A Simple Long-Tailed Recognition Baseline via Vision-Language Model

Exploring the auxiliary learning for long-tailed visual recognition

Contrastive Learning with Hallucinating Data for Long-Tailed Face Recognition.

Open Long-Tailed Recognition In A Dynamic World