Interpretive model of emotional expression in sculptural works based on visual-affective semantic analysis

The expression of emotions in sculptures is conveyed through structural forms, spatial relationships, material characteristics, and symbolic elements. However, traditional artistic analysis mainly relies on expert interpretation, making emotional expression difficult to quantify. With the development of computer vision and multimodal Artificial Intelligence (AI), deep learning approaches provide new opportunities for visual art understanding. Nevertheless, existing methods still face challenges in establishing connections between visual features and abstract emotional semantics, as well as providing interpretable prediction processes. To address these issues, an interpretive model of emotional expression in sculptures based on visual-affective semantic analysis is proposed. A public sculpture image dataset containing 1,800 instances is constructed. A 50-layer Residual Network (ResNet50) is employed to extract morphological and textural features, while Contrastive Language–Image Pre-training (CLIP) is used to establish associations between visual representations and affective semantics. An attention-based fusion mechanism integrates visual and semantic features for emotion classification. Gradient-weighted Class Activation Mapping (Grad-CAM) and semantic ablation methods are further applied to analyze the decision-making process of the model. Experimental results show that the visual-semantic fusion model achieves an accuracy of 82.59% and a Macro-averaged F1 score (Macro-F1) of 80.73% on the test set, outperforming the ResNet50 visual model with 76.30% accuracy and 75.67% Macro-F1, as well as the CLIP semantic model with 73.85% accuracy and 72.58% Macro-F1. Interpretation analysis demonstrates that the model mainly focuses on essential artistic elements, including posture, contour, and spatial structure, with strong consistency with expert evaluations. This research provides an effective artificial intelligence framework for sculptural emotion analysis and offers potential applications in digital museums, art education, intelligent retrieval systems, and computational cultural heritage studies.

Authors

Institutions

Publication Details

Journal
Advances in Humanities Research
Published
2026-11-18
DOI
https://doi.org/10.54254/2753-7080/2026.39245
Primary Topic
Advanced Neural Network Applications
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Interpretive model of emotional expression in sculptural works based on visual-affective semantic analysis

An Zhao
1 links
Advances in Humanities Research
Advanced Neural Network Applications
article

Interpretive model of emotional expression in sculptural works based on visual-affective semantic analysis

An Zhao
article en

Abstract

The expression of emotions in sculptures is conveyed through structural forms, spatial relationships, material characteristics, and symbolic elements. However, traditional artistic analysis mainly relies on expert interpretation, making emotional expression difficult to quantify. With the development of computer vision and multimodal Artificial Intelligence (AI), deep learning approaches provide new opportunities for visual art understanding. Nevertheless, existing methods still face challenges in establishing connections between visual features and abstract emotional semantics, as well as providing interpretable prediction processes. To address these issues, an interpretive model of emotional expression in sculptures based on visual-affective semantic analysis is proposed. A public sculpture image dataset containing 1,800 instances is constructed. A 50-layer Residual Network (ResNet50) is employed to extract morphological and textural features, while Contrastive Language–Image Pre-training (CLIP) is used to establish associations between visual representations and affective semantics. An attention-based fusion mechanism integrates visual and semantic features for emotion classification. Gradient-weighted Class Activation Mapping (Grad-CAM) and semantic ablation methods are further applied to analyze the decision-making process of the model. Experimental results show that the visual-semantic fusion model achieves an accuracy of 82.59% and a Macro-averaged F1 score (Macro-F1) of 80.73% on the test set, outperforming the ResNet50 visual model with 76.30% accuracy and 75.67% Macro-F1, as well as the CLIP semantic model with 73.85% accuracy and 72.58% Macro-F1. Interpretation analysis demonstrates that the model mainly focuses on essential artistic elements, including posture, contour, and spatial structure, with strong consistency with expert evaluations. This research provides an effective artificial intelligence framework for sculptural emotion analysis and offers potential applications in digital museums, art education, intelligent retrieval systems, and computational cultural heritage studies.

Advances in Humanities ResearchVol. 13(12)
Xi'an Jiaotong University (CN)
Advanced Neural Network Applications
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon