Few-shot Class-incremental Learning for Classification and Object Detection: A Survey
It addresses the challenge of incrementally learning new classes from few labeled samples without forgetting, which is crucial for real-world AI applications, but as a survey, it is incremental in summarizing existing work.
This paper provides a comprehensive survey of Few-shot Class-Incremental Learning (FSCIL), covering problem definitions, challenges like unreliable empirical risk minimization and the stability-plasticity dilemma, and methods for classification and object detection.
Few-shot Class-Incremental Learning (FSCIL) presents a unique challenge in Machine Learning (ML), as it necessitates the Incremental Learning (IL) of new classes from sparsely labeled training samples without forgetting previous knowledge. While this field has seen recent progress, it remains an active exploration area. This paper aims to provide a comprehensive and systematic review of FSCIL. In our in-depth examination, we delve into various facets of FSCIL, encompassing the problem definition, the discussion of the primary challenges of unreliable empirical risk minimization and the stability-plasticity dilemma, general schemes, and relevant problems of IL and Few-shot Learning (FSL). Besides, we offer an overview of benchmark datasets and evaluation metrics. Furthermore, we introduce the Few-shot Class-incremental Classification (FSCIC) methods from data-based, structure-based, and optimization-based approaches and the Few-shot Class-incremental Object Detection (FSCIOD) methods from anchor-free and anchor-based approaches. Beyond these, we present several promising research directions within FSCIL that merit further investigation.