Cross-Domain Few-Shot Learning via Adaptive Transformer Networks

TL;DR

This paper introduces ADAPTER, an adaptive transformer network designed for cross-domain few-shot learning, effectively handling large domain shifts by learning transferable features and improving prediction consistency.

Contribution

It proposes a novel adaptive transformer architecture with bidirectional cross-attention and label smoothing, advancing cross-domain few-shot learning capabilities.

Findings

Outperforms prior methods on BSCD-FSL benchmarks

Uses DINO for diverse feature extraction

Achieves significant performance improvements

Abstract

Most few-shot learning works rely on the same domain assumption between the base and the target tasks, hindering their practical applications. This paper proposes an adaptive transformer network (ADAPTER), a simple but effective solution for cross-domain few-shot learning where there exist large domain shifts between the base task and the target task. ADAPTER is built upon the idea of bidirectional cross-attention to learn transferable features between the two domains. The proposed architecture is trained with DINO to produce diverse, and less biased features to avoid the supervision collapse problem. Furthermore, the label smoothing approach is proposed to improve the consistency and reliability of the predictions by also considering the predicted labels of the close samples in the embedding space. The performance of ADAPTER is rigorously evaluated in the BSCD-FSL benchmarks in which it outperforms prior arts with significant margins.