Semantically Proportional Patchmix for Few-Shot Learning

TL;DR

This paper introduces Semantically Proportional Patchmix (SePPMix), a novel data augmentation technique for few-shot learning that improves model generalization by mixing image patches based on semantic content and using rotation prediction as regularization.

Contribution

The paper proposes SePPMix, which mixes image patches proportionally to their semantic information and incorporates rotation prediction to enhance few-shot learning models.

Findings

SePPMix improves few-shot classification accuracy across benchmarks.

Semantic-based patch mixing enhances model generalization.

Rotation regularization further boosts performance.

Abstract

Few-shot learning aims to classify unseen classes with only a limited number of labeled data. Recent works have demonstrated that training models with a simple transfer learning strategy can achieve competitive results in few-shot classification. Although excelling at distinguishing training data, these models are not well generalized to unseen data, probably due to insufficient feature representations on evaluation. To tackle this issue, we propose Semantically Proportional Patchmix (SePPMix), in which patches are cut and pasted among training images and the ground truth labels are mixed proportionally to the semantic information of the patches. In this way, we can improve the generalization ability of the model by regional dropout effect without introducing severe label noise. To learn more robust representations of data, we further take rotate transformation on the mixed images and predict rotations as a rule-based regularizer. Extensive experiments on prevalent few-shot benchmarks have shown the effectiveness of our proposed method.