Image experience prediction for historic districts using a CNN-transformer fusion model
This study addresses key challenges in historic district planning and design: capturing the emotional value of streetscape images and integrating this into the design process. We developed a deep learning-based sentiment analysis system, employing CNN and transformer models to analyze emotional ten...
Saved in:
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Slovenian Society for Stereology and Quantitative Image Analysis
2025-02-01
|
| Series: | Image Analysis and Stereology |
| Subjects: | |
| Online Access: | https://www.ias-iss.org/ojs/IAS/article/view/3361 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | This study addresses key challenges in historic district planning and design: capturing the emotional value of streetscape images and integrating this into the design process. We developed a deep learning-based sentiment analysis system, employing CNN and transformer models to analyze emotional tendencies and temporal states in images. Using a multi-view feature extraction framework combining VGG, ResNet CNNs, and the Swin Transformer model, we created a novel feature matrix. An attention mechanism and transfer learning strategy enhanced model accuracy in label recognition and classification. Applying this system to Jiangnan Historic District, we demonstrated how understanding and applying emotional value can enhance district appeal. By identifying emotional tendencies in streetscape images, designers can make better-informed decisions, fostering positive experiences. Our analysis of images from 12 Jiangnan historic districts showed the system’s efficiency in aligning images with existing imaging libraries, providing valuable references and feedback. The results highlight the practical potential of deep learning in visual sentiment analysis and emphasize the importance of emotional value in improving experiences in historic districts. This study offers new insights and methodological support for planning and designing such areas.
|
|---|---|
| ISSN: | 1580-3139 1854-5165 |