Local Descriptors Weighted Adaptive Threshold Filtering For Few-Shot Learning

TL;DR

This work tackles the challenge of background noise in local descriptors for $N$-way $K$-shot few-shot learning by introducing a weighted adaptive threshold filtering (WATF) framework. The method comprises EFEM for local descriptor extraction, WATF to dynamically filter category-relevant descriptors using prototype-based weights and an adaptive threshold, and KLDCM to classify filtered descriptors via a $k$-NN–based score with softmax. Key contributions include a lightweight, parameter-free filtering strategy that improves clustering of descriptor features and enhances cross-class discrimination, achieving state-of-the-art results on CUB-200, Stanford Dogs, and Stanford Cars, plus strong cross-domain performance on miniImageNet$\rightarrow$CUB. The approach yields practical impact by delivering robust few-shot performance with reduced computational overhead and potential applicability beyond imaging to other modalities such as medical imaging and text.

Abstract

Few-shot image classification is a challenging task in the field of machine learning, involving the identification of new categories using a limited number of labeled samples. In recent years, methods based on local descriptors have made significant progress in this area. However, the key to improving classification accuracy lies in effectively filtering background noise and accurately selecting critical local descriptors highly relevant to image category information. To address this challenge, we propose an innovative weighted adaptive threshold filtering (WATF) strategy for local descriptors. This strategy can dynamically adjust based on the current task and image context, thereby selecting local descriptors most relevant to the image category. This enables the model to better focus on category-related information while effectively mitigating interference from irrelevant background regions. To evaluate the effectiveness of our method, we adopted the N-way K-shot experimental framework. Experimental results show that our method not only improves the clustering effect of selected local descriptors but also significantly enhances the discriminative ability between image categories. Notably, our method maintains a simple and lightweight design philosophy without introducing additional learnable parameters. This feature ensures consistency in filtering capability during both training and testing phases, further enhancing the reliability and practicality of the method.

Figures (6)

• Figure 1: Examples of regions that are relevant and irrelevant to image classes.
• Figure 2: Visualization of probability distribution histogram of the average weights of local descriptors for support set images.
• Figure 3: Visualization results of local descriptors for four randomly sampled 5-way 1-shot classification tasks, comparing the cases with and without our WATF module.
• Figure 4: Training loss, validation loss, and test loss curves of the proposed WATF on 5-way 1-shot setting of three datasets.
• Figure 5: Visualization of features before and after applying the WATF operation on three datasets.