Table of Contents
Fetching ...

Feature Aligning Few shot Learning Method Using Local Descriptors Weighted Rules

Bingchen Yan

TL;DR

FAFD-LDWR tackles the noise and background interference inherent in local-descriptor–based few-shot learning by introducing cross normalization to preserve discriminative information and a dynamic, neighborhood-based weighting scheme to filter descriptors. The method leverages neighborhood representations of local descriptors, constructs class prototypes from the supports, and classifies queries using filtered descriptors through an image-to-class cosine matching framework. Across three benchmarks (CUB-200, Stanford Dogs, Stanford Cars) and both Conv4 and ResNet-12 backbones, FAFD-LDWR achieves state‑of‑the‑art performance in most 5-way 1-shot and 5-shot settings, with notable gains when using deeper backbones; cross-domain evaluation on miniImageNet→CUB also demonstrates strong generalization. The work demonstrates improved robustness and interpretability by focusing on salient local regions and can extend to other data modalities, offering a practical advance for real-world few-shot recognition tasks.

Abstract

Few-shot classification involves identifying new categories using a limited number of labeled samples. Current few-shot classification methods based on local descriptors primarily leverage underlying consistent features across visible and invisible classes, facing challenges including redundant neighboring information, noisy representations, and limited interpretability. This paper proposes a Feature Aligning Few-shot Learning Method Using Local Descriptors Weighted Rules (FAFD-LDWR). It innovatively introduces a cross-normalization method into few-shot image classification to preserve the discriminative information of local descriptors as much as possible; and enhances classification performance by aligning key local descriptors of support and query sets to remove background noise. FAFD-LDWR performs excellently on three benchmark datasets , outperforming state-of-the-art methods in both 1-shot and 5-shot settings. The designed visualization experiments also demonstrate FAFD-LDWR's improvement in prediction interpretability.

Feature Aligning Few shot Learning Method Using Local Descriptors Weighted Rules

TL;DR

FAFD-LDWR tackles the noise and background interference inherent in local-descriptor–based few-shot learning by introducing cross normalization to preserve discriminative information and a dynamic, neighborhood-based weighting scheme to filter descriptors. The method leverages neighborhood representations of local descriptors, constructs class prototypes from the supports, and classifies queries using filtered descriptors through an image-to-class cosine matching framework. Across three benchmarks (CUB-200, Stanford Dogs, Stanford Cars) and both Conv4 and ResNet-12 backbones, FAFD-LDWR achieves state‑of‑the‑art performance in most 5-way 1-shot and 5-shot settings, with notable gains when using deeper backbones; cross-domain evaluation on miniImageNet→CUB also demonstrates strong generalization. The work demonstrates improved robustness and interpretability by focusing on salient local regions and can extend to other data modalities, offering a practical advance for real-world few-shot recognition tasks.

Abstract

Few-shot classification involves identifying new categories using a limited number of labeled samples. Current few-shot classification methods based on local descriptors primarily leverage underlying consistent features across visible and invisible classes, facing challenges including redundant neighboring information, noisy representations, and limited interpretability. This paper proposes a Feature Aligning Few-shot Learning Method Using Local Descriptors Weighted Rules (FAFD-LDWR). It innovatively introduces a cross-normalization method into few-shot image classification to preserve the discriminative information of local descriptors as much as possible; and enhances classification performance by aligning key local descriptors of support and query sets to remove background noise. FAFD-LDWR performs excellently on three benchmark datasets , outperforming state-of-the-art methods in both 1-shot and 5-shot settings. The designed visualization experiments also demonstrate FAFD-LDWR's improvement in prediction interpretability.
Paper Structure (23 sections, 17 equations, 3 figures, 4 tables)

This paper contains 23 sections, 17 equations, 3 figures, 4 tables.

Figures (3)

  • Figure 1: Examples Of Regions That Are Relevant And Irrelevant To Image Classes.
  • Figure 2: The proposed FAFD-LDWR method's framework for 5-way 1-shot classification.
  • Figure 3: Accuracy As A Function Of Local Descriptor Neighborhood Representation $k$ Value.