Enhanced automated art curation using supervised modified CNN for art style classification.

This study explores the application of a supervised Modified Convolutional Neural Network (CNN) for automated art classification and curation. Traditional art classification methods rely heavily on human expertise, which is time-consuming, subjective, and inconsistent. To address these challenges, we developed a Modified CNN model capable of distinguishing art styles and movements using features such as color patterns, textures, and compositions. The model was trained and evaluated on a custom dataset comprising 5000 artworks representing five major art styles: Impressionism, Cubism, Realism, Abstract, and Surrealism. The Modified CNN achieved an average classification accuracy of 93.0%, surpassing existing models such as ResNet50 and VGG16 in precision (93.5%), recall (92.8%), and F1-score (93.1%). Feature visualization using t-SNE and PCA highlighted the model's ability to cluster distinct styles while identifying overlaps in challenging categories such as Abstract and Surrealism. Grad-CAM heatmaps provided insights into regions contributing to incorrect predictions, revealing opportunities for refinement. Despite its strong performance, the model faced limitations, including biases in training data and overlapping stylistic features. Future work aims to expand datasets, incorporate multimodal inputs, and improve interpretability using explainable AI techniques. This research demonstrates the potential of Modified CNNs as a scalable and consistent tool for art classification, with applications in digital curation, art education, and cultural preservation.