Rescaled Hinge Loss Function based Convolutional Neural Network for Identification of Art Forgery by Image Analysis

The rapid growth of Information Technology (IT) as well as Computer Science becomes an important evolution in marketable painting applications. The systematic authentication of paintings embraces the implication into the monarchy of art gathering. Performing the Convolutional Neural Network (CNN) approach for the identification of the reliable as well as forged painting arts according to the colour images is a feasible but trivial selection. However, this is challenging due to variations in artistic styles, textures, and subtle details which ambiguous forgery detection. Hence, this research proposes the Rescaled Hinge Loss Function based CNN (RHLF-CNN) approach for the identification of art forgery. The RHLF improves the model’s capability to discriminate between genuine and forged artworks, leads to more precise classification. The high-definition and hyperspectral image dataset of Chinese characters is utilized to estimate the effectiveness of the RHLF-CNN approach. The Local Binary Pattern (LBP) approach is used for the feature extraction process. The Chi-square approach is utilized for extracting the relevant features from the extraction process. The experimental results demonstrates that the proposed RHLF-CNN approach attains the better accuracy of 93.16% as compared to the existing approaches like Pulse-Coupled Neural Network (PCNN).