Improved Embedded Learning for Classification of Chinese Paintings

The existing research on classification of Chinese paintings was limited to consideration of the relationship between paintings and labels,where the distribution information of paintings and feature importance were often ignored.Firstly,we extracted the features of Chinese paintings by fine-tuning VGG-F models.Then,the mutual information theory was introduced into the embedded machine learning,and formulated a new embedded classification algorithm to characterize the variation in artistic styling of the Chinese paintings.Finally,support vector machine was applied as the classifier to obtain satisfactory classification results.The average classification accuracy of 10 artists Chinese paintings is 86%.Experimental results support that the proposed algorithm outperforms the existing representative benchmarks.