Interpretable Modeling of Aesthetic Evaluation in Calligraphy Imitation
Authors:
- 30 June 2025
- conference paper
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-10
Abstract
Calligraphy is a brilliant gem of Chinese culture, and achieving a calligraphy imitation aesthetic evaluation that transcends human judgment while maintaining interpretability is of great significance for the inheritance of this art. This study first constructs a large-scale calligraphy imitation dataset consisting of 605 template characters, 191 participants, and 48,826 imitation characters. Then, 48-dimensional interpretable calligraphic aesthetic features are designed. By integrating the 48-dimensional calligraphic aesthetic features with deep learning features to varying extents, four calligraphy imitation evaluation models are constructed. Subsequently, the SHAP method is used for the interpretability evaluation of each model, analyzing the contribution of each feature, ultimately resulting in a model that not only surpasses human judgment but also possesses high interpretability. The interpretability analysis results indicate that, when features are ranked by contribution, the model only requires the top 23 aesthetic features to outperform human judgment in predictive effectiveness.Keywords
This publication has 32 references indexed in Scilit:
- SRAFE: Siamese Regression Aesthetic Fusion Evaluation for Chinese Calligraphic CopyCAAI Transactions on Intelligence Technology, 2022
- Constructing Calligraphy Evaluation Model Based on Writing Movement with LSTM Network2022
- Ensemble of handcrafted and deep features for urban sound classificationApplied Acoustics, 2021
- Inkthetics: A Comprehensive Computational Model for Aesthetic Evaluation of Chinese Ink PaintingsIEEE Access, 2020
- Deep imitator: Handwriting calligraphy imitation via deep attention networksPattern Recognition, 2020
- Chinese calligraphic style representation for recognitionInternational Journal on Document Analysis and Recognition (IJDAR), 2017
- Deep Residual Learning for Image Recognition2016
- Simulation of Human Opinions about Calligraphy Aesthetic2014
- Computationally evaluating and synthesizing Chinese calligraphyNeurocomputing, 2014
- An interactive grading and learning system for chinese calligraphyMachine Vision and Applications, 2007