To address the problem of imprecise image classification of Chinese paintings due to insufficient feature extraction, this paper proposed an image classification of Chinese paintings based on attention-aware deep feature embedding. Firstly, the Swin-Transformer framework is utilized to construct the backbone network for extracting features from images. Secondly, the dual-attention mechanism is constructed to acquire global semantic features and local object features. Finally, the extracted features are fused and input into a softmax classifier for image classification. To verify the feasibility and effectiveness of the proposed image classification model for Chinese paintings, simulation experiments are designed based on the Chinese Painting Datase. Moreover, comparative experiments are conducted against existing advanced and representative algorithms to demonstrate that the proposed model outperforms the existing models in terms of both visual performance and objective data comparison.