Generative AI for architectural heritage conservation: expert-guided feature encoding, sty

A dataset of 148 annotated images of Chinese timber architectural elements, used for AI-generated content research. The dataset was created by Qianru Yang and last updated on June 1, 2026. It focuses on the Qiqushan Great Temple in Sichuan and includes four element categories: eave decorations, arch structures, interior beams, and pattern motifs.

Use Cases

• Fine-tuning generative models for architectural element synthesis based on annotated image categories.
• Evaluating AI-generated content control parameters for heritage conservation based on the described Edge Alignment Score analysis.
• Exploring stylistic regeneration of traditional Chinese timber building components like Dougong based on the expert-guided workflow.

Strengths

• Dataset contains 148 verified and annotated images.
• Analysis includes evaluation across nine control weights (s=0.1–1.5) with an identified optimal range.
• Workflow integrates multiple AIGC tools (ChatGPT-5.1, Midjourney v6.1, Stable Diffusion 3.5) for a specific case study.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• The dataset is relatively small at 5.4 MB.

Provenance

Source

Author: Qianru Yang. Platform: figshare.

Collection Method

Likely created through expert annotation of images from the Qiqushan Great Temple case study.

Freshness

Last updated 2026-06-01 05:02:13; freshness should be verified.

Geography

Qiqushan Great Temple in Sichuan, China.