Semantic Regularization: Improve Few-shot Image Classification by Reducing Meta Shift

TL;DR

This paper introduces a Semantic Regularization Network that reduces meta shift in few-shot image classification by learning a shared semantic space, leading to more stable class descriptors and improved performance.

Contribution

The paper proposes a novel semantic regularization framework within meta-learning to address meta shift, enhancing class descriptor stability in few-shot learning.

Findings

Improves accuracy by 4%-7% over baseline methods.

Achieves competitive results on MiniImageNet, TieredImageNet, and CUB datasets.

Effectively reduces meta shift across multiple tasks.

Abstract

Few-shot image classification requires the classifier to robustly cope with unseen classes even if there are only a few samples for each class. Recent advances benefit from the meta-learning process where episodic tasks are formed to train a model that can adapt to class change. However, these task sare independent to each other and existing works mainly rely on limited samples of individual support set in a single meta task. This strategy leads to severe meta shift issues across multiple tasks, meaning the learned prototypes or class descriptors are not stable as each task only involves their own support set. To avoid this problem, we propose a concise Semantic RegularizationNetwork to learn a common semantic space under the framework of meta-learning. In this space, all class descriptors can be regularized by the learned semantic basis, which can effectively solve the meta shift problem. The key is to train a class encoder and decoder structure that can encode the sample embedding features into the semantic domain with trained semantic basis, and generate a more stable and general class descriptor from the decoder. We evaluate our work by extensive comparisons with previous methods on three benchmark datasets (MiniImageNet, TieredImageNet, and CUB). The results show that the semantic regularization module improves performance by 4%-7% over the baseline method, and achieves competitive results over the current state-of-the-art models.