Cross-domain few-shot learning via adaptive transformer networks

Naeem Paeedeh, Mahardhika Pratama, Muhammad Anwar Ma'sum, Wolfgang Mayer, Zehong Cao, Ryszard Kowlczyk

Research output: Contribution to journalArticlepeer-review

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.

Original languageEnglish
Article number111458
JournalKnowledge-Based Systems
Volume288
DOIs
Publication statusPublished - 15 Mar 2024

Keywords

  • Cross-domain few-shot learning
  • Domain adaptation
  • Few-shot learning

Fingerprint

Dive into the research topics of 'Cross-domain few-shot learning via adaptive transformer networks'. Together they form a unique fingerprint.

Cite this