Artistic Style Classification using Convolution Sparse Coding

Artistic style classification is an important part of visual artwork study. This paper describes a new approach for enhancing the accuracy of artist classification for line drawing images. We rely on the use of convolutional sparse coding that is known for its ability to model artistic styles through convolutional dictionary elements that represent artist brush strokes. We evaluate this framework by using two different datasets formed of various line drawing images generated by multiple artists where each artist owns a different artistic style. Once the dataset is prepared, it is preprocessed, and then used to train a convolutional sparse coding model to distinguish the artistic styles. This facilitates the process of discriminative feature learning of artistic styles by relying on the process of convolutional dictionary learning. The experimental results show that our approach outperforms other alternatives and guarantees significant progress for enhancing the field of artist style classification.