A Closer Look at Few-shot Classification

TLDR

Few-shot classification seeks to learn classifiers that recognize unseen classes from limited labeled examples, yet the increasing complexity of network designs, meta-learning algorithms, and implementation variations hampers fair comparison. In this paper, we present a consistent comparative analysis of several representative few-shot classification algorithms, a modified baseline method that surprisingly achieves competitive performance, and a new experimental setting for evaluating cross-domain generalization. The authors compare several representative few-shot classification algorithms using deeper backbones to reduce performance differences, introduce a modified baseline method, and establish a new cross-domain evaluation setting. Our results show that reducing intra-class variation is crucial for shallow backbones but less so for deeper ones, and that a baseline method with standard fine-tuning performs favorably against state‑of‑the‑art algorithms in realistic cross-domain evaluations.

Abstract

Few-shot classification aims to learn a classifier to recognize unseen classes during training with limited labeled examples. While significant progress has been made, the growing complexity of network designs, meta-learning algorithms, and differences in implementation details make a fair comparison difficult. In this paper, we present 1) a consistent comparative analysis of several representative few-shot classification algorithms, with results showing that deeper backbones significantly reduce the performance differences among methods on datasets with limited domain differences, 2) a modified baseline method that surprisingly achieves competitive performance when compared with the state-of-the-art on both the \miniI and the CUB datasets, and 3) a new experimental setting for evaluating the cross-domain generalization ability for few-shot classification algorithms. Our results reveal that reducing intra-class variation is an important factor when the feature backbone is shallow, but not as critical when using deeper backbones. In a realistic cross-domain evaluation setting, we show that a baseline method with a standard fine-tuning practice compares favorably against other state-of-the-art few-shot learning algorithms.