Enhanced Art Style Classification with VGG16 and Machine Learning Techniques

Art is an expression of human creativity and culture that has greatly impacted history. Classifying art styles is an important process in art and culture to preserve cultural heritage and facilitate educational and research purposes. The proposed work presents a comprehensive study that focuses on classifying 13 different art styles using deep learning techniques. The VGG16 convolutional neural network was employed for feature extraction using its convolutional and pooling layers up to the ’fc1’ layer, generating 512 and 4096-dimensional feature vectors. To evaluate the classification performance, seven machine learning classifiers were utilized. The classification accuracy is evaluated before and after applying Principal Component Analysis (PCA), to reduce the dimensionality of the feature vectors. Additionally, the paper explored the direct classification by the VGG16 model. Results prove that deep feature learning with the combination of traditional classifiers of machine learning is efficient.