Cross-domain Few-shot Hyperspectral Image Classification With Class-wise Attention

• Published in: IEEE transactions on geoscience and remote sensing 2023-01, Vol.61, p.1-1
• Main Authors: Wang, Wenzhen, Liu, Fang, Liu, Jia, Xiao, Liang
• Format: Article
• Language: English
• Subjects: Asymmetry; class-wise attention; Classification; cross-domain; Data mining; Feature extraction; few-shot learning; hyperspectral image classification; Hyperspectral imaging; Image classification; Measurement; Methods; Modules; Prototypes; Task analysis; Training
• ISSN: 0196-2892; 1558-0644
• DOI: 10.1109/TGRS.2023.3239411

Few-shot learning (FSL) is an effective method to solve the problem of hyperspectral image classification with few labeled samples. It learns transferable knowledge from sufficient labeled auxiliary data to classify unseen classes with limited labeled samples for training. However, the distribution difference between auxiliary data and unseen classes results in the learned transferable knowledge not being well applied to the new task. Therefore, a class-wise attentive cross-domain few-shot learning (CA-CFSL) framework is proposed in this paper, in which a feature extractor is learned to extract data features with discriminability and domain invariance. The class-wise attention metric module (CAMM) introduces a class-wise attention on the FSL framework to learn more discriminative features, which improves the inter-class decision boundaries. Furthermore, an asymmetric domain adversarial module (ADAM) is designed to enhance the ability of extracting domain invariant representations, which combines asymmetric adversarial training with embedded domain-specific information. Experimental results on four public hyperspectral image datasets demonstrate that the proposed method outperforms existing methods.