Using ensemble learning for classifying artistic styles in traditional Chinese woodcuts

As one of the oldest art fields, Chinese woodcut represent the rich history and culture of China, and have so far seen various artistic styles and diverse techniques. However, an automated method for identifying and classifying the artistic styles of these works has not yet been proposed. In this work, we propose a new method that consists of three main steps: The first step involves the compilation and preprocessing of the images of the woodblocks and their conversion into colored form to extract important features. The second step will use an ensemble model based on Convolutional Neural Network (CNN), which allows for the prediction of the time period and artistic style of the works. In the third step, the predicted results from the basic models are combined with a regression tree-based meta-model Classification and Regression Tree (CART) to improve the prediction accuracy. This hybrid and multi-stage approach seeks to provide an efficient solution for the automatic identification and classification of Chinese woodcut and can contribute to scientific and applied advances in the field of Chinese art. The results of experiments conducted on our proposed method show remarkable success in achieving high levels of accuracy and precision in the identification and classification of Chinese woodcut. Specifically, this method has been able to achieve an accuracy increase of 4.7% and a precision increase level of 4%, which represent significant improvements compared to the compared methods. These results demonstrate the model's ability to more accurately and efficiently identify different artistic styles and time periods in woodcut.