Few-Shot Fine-Grained Image Recognition via Pairwise Alignment and Progressive Competence Training
Journal of the Artificial Intelligence Engineers Consortium · 2026
DOI: 10.66238/jaiec55 ↗Kurz gefasst
Fachartikel in der Zeitschrift Journal of the Artificial Intelligence Engineers Consortium, 2026. Verfasst von Valeria Lombardi und 1 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
The intersection of few-shot learning and fine-grained image recognition presents a formidable challenge in contemporary computer vision. Models must learn to distinguish highly similar sub-categories, such as different species of birds or models of aircraft, using only a minimal number of annotated training examples. Current approaches often struggle with spatial misalignment of discriminative parts and the inherent difficulty of optimizing complex models on sparse data distributions. To address these critical limitations, this paper proposes a novel framework centered on Pairwise Alignment and Progressive Competence Training. The Pairwise Alignment module focuses on establishing dense semantic correspondences between the local descriptors of query and support images, effectively mitigating background noise and spatial variations without requiring part-level annotations. Concurrently, the Progressive Competence Training strategy introduces a curriculum-based learning paradigm. By dynamically estimating the difficulty of recognition tasks based on feature space density and inter-class margins, the model is trained sequentially from broadly distinguishable categories to highly nuanced fine-grained tasks. This prevents premature convergence and stabilizes the meta-learning process. Extensive evaluations across multiple standard benchmarks demonstrate that the proposed framework significantly outperforms existing state-of-the-art methods in both one-shot and five-shot settings. The descriptive analysis elucidates the contribution of each module, confirming that combining explicit part alignment with structured learning trajectories yields a highly robust and generalization-capable architecture for low-data visual recognition tasks.
Zitieren
APA
Lombardi, V., & Klein, H. (2026). Few-Shot Fine-Grained Image Recognition via Pairwise Alignment and Progressive Competence Training. Journal of the Artificial Intelligence Engineers Consortium. https://doi.org/10.66238/jaiec55
Vancouver
Lombardi V, Klein H. Few-Shot Fine-Grained Image Recognition via Pairwise Alignment and Progressive Competence Training. Journal of the Artificial Intelligence Engineers Consortium. 2026. doi:10.66238/jaiec55
BibTeX
@article{Lombardi2026,
author = {Lombardi, Valeria and Klein, Henrik},
title = {Few-Shot Fine-Grained Image Recognition via Pairwise Alignment and Progressive Competence Training},
journal = {Journal of the Artificial Intelligence Engineers Consortium},
year = {2026},
doi = {10.66238/jaiec55}
}
EUVIMED