Few-shot fine-grained image classification (FSFGIC) methods refer to the classification of images (e.g., birds, flowers, and airplanes) belonging to different subclasses of the same species by a small number of labeled samples. Through feature representation learning, FSFGIC methods can make better use of limited sample information, learn more discriminative feature representations, greatly improve the classification accuracy and generalization ability, and thus achieve better results in FSFGIC tasks. In this paper, starting from the definition of FSFGIC, a taxonomy of feature representation learning for FSFGIC is proposed. According to this taxonomy, we discuss key issues on FSFGIC (including data augmentation, local and/or global deep feature representation learning, class representation learning, and task-specific feature representation learning). In addition, the existing popular datasets, current challenges and future development trends of feature representation learning on FSFGIC are also described.

Few-shot fine-grained image classification (FSFGIC) methods refer to machine learning methods which aim to classify images (e.g., bird species, flowers, and airplanes) belonging to subordinate object categories of the same entry-level category with only a few samples. It is worth to note that feature representation learning is used not only to represent training samples, but also to construct classifiers for performing various FSFGIC tasks. In this paper, starting from the definition of FSFGIC, a taxonomy of feature representation learning for FSFGIC is proposed. According to this taxonomy, we discuss key issues on FSFGIC (including data augmentation, local or/and global deep feature representation learning, class representation learning, and task specific feature representation learning). The existing popular datasets and evaluation standards are introduced. Furthermore, a novel classification performance evaluation mechanism is designed with a 0.95 confidence interval for judging whether the classification accuracy obtained by a certain specified method is good or bad. Moreover, current challenges and future trends of feature representation learning on FSFGIC are elaborated.

The challenge of few-shot fine-grained image classification lies in the scarcity of labeled training samples, which hinders models from learning subtle discriminative patterns. Moreover, most existing methods rely heavily on spatial-domain convolutional operations while overlooking structured spectral information in the frequency domain, leading to suboptimal representations. To address this limitation, we propose a Multi-Scale Spatial-Frequency Feature Enhancement Network. In the spatial domain, a dual-branch architecture is designed to model both local fine-grained details and long-range contextual dependencies. In the frequency domain, discriminative information is enhanced through multi-scale frequency decomposition and an amplitude-based attention mechanism. The spatial and frequency features are then adaptively fused into a unified representation that integrates complementary information from both domains. Extensive experiments on three public benchmarks, namely CUB-200-2011, Stanford-Dogs, and Stanford-Cars, demonstrate the effectiveness of the proposed framework. The proposed method consistently outperforms state-of-the-art approaches under both 1-shot and 5-shot settings, showing strong discriminative ability and robust generalization performance. The code is available at https://github.com/WW-YM/MSSFFF-Net/tree/master.