Article: Research on fine-grained classification algorithm of oil painting schools based on multilevel SVM and feature engineering Journal: International Journal of Information and Communication Technology (IJICT) 2026 Vol.27 No.1 pp.77 - 99 Abstract: This paper aims to address these issues by proposing a fine-grained classification algorithm system that integrates multi-level support vector machines (SVMs) and feature engineering. This system combines data pre-processing, feature enhancement, and multi-level SVM classification to construct a hierarchical decision-making structure of genre clusters → genres → periods. Experiments show that the algorithm achieves an accuracy of 92.3% on a diverse dataset, with a macro F1 score of 0.915. Furthermore, in robustness tests, the accuracy only drops to 88.0% under noise perturbation, 87.3% under blur perturbation, and 85.5% under occlusion perturbation. The cross-dataset generalisation accuracy reaches 85.2%. External validation indicates an average accuracy of 82.3% for non-Western oil paintings, while also demonstrating high interpretability. This paper contributes an innovative technical approach for fine-grained classification of oil painting genres; it enhances accuracy, robustness, and interpretability, and lays the foundation for subsequent lightweight design, multimodal fusion, and cross-media expansion research. Inderscience Publishers - linking academia, business and industry through research

Title: Research on fine-grained classification algorithm of oil painting schools based on multilevel SVM and feature engineering

Authors: Qi Xie; Chaobin Wang

Addresses: Academy of Painting, Hubei Institute of Fine Arts, Wuhan, Hubei, 430205, China ' Academy of Painting, Hubei Institute of Fine Arts, Wuhan, Hubei, 430205, China

Abstract: This paper aims to address these issues by proposing a fine-grained classification algorithm system that integrates multi-level support vector machines (SVMs) and feature engineering. This system combines data pre-processing, feature enhancement, and multi-level SVM classification to construct a hierarchical decision-making structure of genre clusters → genres → periods. Experiments show that the algorithm achieves an accuracy of 92.3% on a diverse dataset, with a macro F1 score of 0.915. Furthermore, in robustness tests, the accuracy only drops to 88.0% under noise perturbation, 87.3% under blur perturbation, and 85.5% under occlusion perturbation. The cross-dataset generalisation accuracy reaches 85.2%. External validation indicates an average accuracy of 82.3% for non-Western oil paintings, while also demonstrating high interpretability. This paper contributes an innovative technical approach for fine-grained classification of oil painting genres; it enhances accuracy, robustness, and interpretability, and lays the foundation for subsequent lightweight design, multimodal fusion, and cross-media expansion research.

Keywords: multi-level SVM; support vector machines; feature engineering; oil painting; genre; classification.

DOI: 10.1504/IJICT.2026.151310

International Journal of Information and Communication Technology, 2026 Vol.27 No.1, pp.77 - 99

Received: 29 Sep 2025
Accepted: 16 Nov 2025
Published online: 22 Jan 2026 *

Title: Research on fine-grained classification algorithm of oil painting schools based on multilevel SVM and feature engineering

Keep up-to-date