Transductive Few-Shot Classification on the Oblique Manifold

TL;DR

This paper introduces a novel transductive few-shot classification method on the Oblique Manifold, combining a new feature extraction technique, a specialized classifier, and transductive learning strategies, achieving superior results on standard benchmarks.

Contribution

It proposes a new approach for few-shot learning on the Oblique Manifold using a non-parametric attention mechanism and a tangent space classifier, with improved initialization and loss functions.

Findings

Outperforms state-of-the-art on mini-ImageNet, tiered-ImageNet, and CUB datasets.

Effective feature extraction with RSSPP improves discriminative ability.

The Oblique Distance-based Classifier enhances local approximation of the manifold.

Abstract

Few-shot learning (FSL) attempts to learn with limited data. In this work, we perform the feature extraction in the Euclidean space and the geodesic distance metric on the Oblique Manifold (OM). Specially, for better feature extraction, we propose a non-parametric Region Self-attention with Spatial Pyramid Pooling (RSSPP), which realizes a trade-off between the generalization and the discriminative ability of the single image feature. Then, we embed the feature to OM as a point. Furthermore, we design an Oblique Distance-based Classifier (ODC) that achieves classification in the tangent spaces which better approximate OM locally by learnable tangency points. Finally, we introduce a new method for parameters initialization and a novel loss function in the transductive settings. Extensive experiments demonstrate the effectiveness of our algorithm and it outperforms state-of-the-art methods on the popular benchmarks: mini-ImageNet, tiered-ImageNet, and Caltech-UCSD Birds-200-2011 (CUB).