Object-Level Representation Learning for Few-Shot Image Classification

TL;DR

This paper introduces an object-level relation approach leveraging an auxiliary dataset to enhance few-shot image classification accuracy without fine-tuning, demonstrating significant improvements on Omniglot and MiniImagenet datasets.

Contribution

It proposes a novel object-level relation learning method that utilizes additional datasets to improve few-shot classification performance without requiring model fine-tuning.

Findings

Achieved 8.5% and 2.7% accuracy improvements on MiniImagenet for 5-way 1-shot and 5-shot tasks.

Utilized object-level relations from auxiliary datasets to infer image similarities.

Method does not require fine-tuning due to non-parametric nearest neighbor classification.

Abstract

Few-shot learning that trains image classifiers over few labeled examples per category is a challenging task. In this paper, we propose to exploit an additional big dataset with different categories to improve the accuracy of few-shot learning over our target dataset. Our approach is based on the observation that images can be decomposed into objects, which may appear in images from both the additional dataset and our target dataset. We use the object-level relation learned from the additional dataset to infer the similarity of images in our target dataset with unseen categories. Nearest neighbor search is applied to do image classification, which is a non-parametric model and thus does not need fine-tuning. We evaluate our algorithm on two popular datasets, namely Omniglot and MiniImagenet. We obtain 8.5\% and 2.7\% absolute improvements for 5-way 1-shot and 5-way 5-shot experiments on MiniImagenet, respectively. Source code will be published upon acceptance.