Layer-Wise Feature Metric of Semantic-Pixel Matching for Few-Shot Learning

TL;DR

This paper introduces LWFM-SPM, a novel approach for few-shot learning that improves semantic pixel matching by layer-wise feature comparison and semantic alignment, addressing spatial misalignment issues in natural scenes.

Contribution

The paper proposes a new method combining layer-wise embedding and semantic-pixel matching modules to enhance similarity measurement in few-shot learning.

Findings

Achieves competitive results on miniImageNet, tieredImageNet, CUB-200-2011, and CIFAR-FS.

Addresses spatial misalignment in semantic pixel matching.

Provides a publicly available implementation.

Abstract

In Few-Shot Learning (FSL), traditional metric-based approaches often rely on global metrics to compute similarity. However, in natural scenes, the spatial arrangement of key instances is often inconsistent across images. This spatial misalignment can result in mismatched semantic pixels, leading to inaccurate similarity measurements. To address this issue, we propose a novel method called the Layer-Wise Features Metric of Semantic-Pixel Matching (LWFM-SPM) to make finer comparisons. Our method enhances model performance through two key modules: (1) the Layer-Wise Embedding (LWE) Module, which refines the cross-correlation of image pairs to generate well-focused feature maps for each layer; (2)the Semantic-Pixel Matching (SPM) Module, which aligns critical pixels based on semantic embeddings using an assignment algorithm. We conducted extensive experiments to evaluate our method on four widely used few-shot classification benchmarks: miniImageNet, tieredImageNet, CUB-200-2011, and CIFAR-FS. The results indicate that LWFM-SPM achieves competitive performance across these benchmarks. Our code will be publicly available on https://github.com/Halo2Tang/Code-for-LWFM-SPM.