Abstract: Low-resolution (LR) fine-grained image recognition requires the ability to recognize the subcategories of LR samples with limited fine-grained details. The existing methods do not make full use of the guiding and constraining capabilities of category-related knowledge to recover and extract the fine-grained features of LR data; thus these methods have a limited ability to learn the global and local fine-grained features of LR data. In this paper, we propose an enhanced feature representation network (EFR-Net) based on categorical knowledge guidance to capture delicate and reliable fine-grained feature descriptions of LR data and improve the recognition accuracy. First, to overcome the challenges posed by the limited fine-grained details in LR data, we design a classwise distillation loss. This loss function transfers the high-quality features of class-specific high-resolution (HR) samples into the feature learning of the same-category LR samples by using a memory bank. In this way, the global representation of LR images is closer to the meaningful and high-quality image features. Second, considering that fine-grained discriminative features are often hidden in object parts, we present a group of part queries to learn the positional information where the discriminative cues exist across all categories, and we then use the queries to decode diverse and discriminative part features. The global representation, in combination with the local discriminative features, creates more comprehensive and meaningful feature descriptions of the LR fine-grained objects, thus improving the recognition performance. Extensive comparison experiments on four LR datasets demonstrate the effectiveness of EFR-Net.