The conservation of traditional villages has shifted from isolated site-by-site protection to regional collaboration, and exploring pathways for their sustainable development has become a key focus of research. Existing research still falls short in areas such as the integration of heritage value into decision-making mechanisms and the establishment of systematic conservation frameworks, leading to prominent issues of isolated conservation and homogeneous development. Taking traditional villages in Yanchuan County, China, as a case study, this research aims to establish a clustered conservation system and achieve a transition towards networked collaborative governance. The study utilised field surveys and literature review to establish a database and systematically catalogue heritage resources; it combined the Analytic Hierarchy Process (AHP) and the Delphi method to construct a value evaluation system and identify distinctive features; and it integrated cluster theory with GIS spatial analysis to construct a clustered conservation framework across three dimensions: classification and grading, symbiotic models, and the overall spatial pattern. The results indicate that: (1) the spatial distribution of villages in Yanchuan County is uneven, and the villages themselves exhibit significant homogeneity in their characteristics; (2) core characteristics include Loess culture, cave dwellings and revolutionary heritage sites, with comprehensive scores ranging from 0.4437 to 0.9116; these are classified into three protection levels, identifying five categories of villages of value. (3) Five major cluster zones were delineated based on resource and spatial characteristics. By integrating river basins and transport corridors, a comprehensive protection framework of ‘one belt, two wings, two centers and five zones’ was established, alongside three types of cluster symbiosis models, thereby achieving regional resource integration and enhancing collaborative efficiency. The cluster-based protection system proposed in this study can effectively address the challenges facing the conservation and development of traditional villages, providing a feasible solution for regional collaborative protection, and holds practical significance for cultural heritage management and sustainable development.

Cultural heritage value serves as a critical foundation for guiding the protection and utilization of cultural heritage. However, existing methods for evaluating cultural heritage value face several inherent limitations, including evaluator subjectivity, high resource demands, and a lack of timely and objective evaluation capabilities. To overcome these challenges, we present the first application of deep learning to heritage value evaluation through our proposed Vision Transformer-based Heritage Value Evaluation model (ViT-HVE). This novel framework represents a breakthrough in automated cultural heritage analysis, enabling both visual recognition and quantitative evaluation of heritage values through an end-to-end deep learning approach. In this paper, the cultural heritage of the Yellow River in Shaanxi province is selected as a case study. A ten-dimensional value system characterizing these cultural heritage is extracted using the Latent Dirichlet Allocation (LDA) topic model. Based on this, we developed the first Cultural Heritage Value Recognition (CHVR) dataset in this domain. To facilitate the quantitative evaluation of cultural heritage values, we further introduce a Top-k Heritage Value Weighting (TK-HVW) method, which performs value weighting by extracting and normalizing the probability distributions over heritage value categories. To evaluate the effectiveness of the proposed method, we conduct a comparative analysis between ViT-HVE and five state-of-the-art deep learning models, and validate the accuracy of the model’s weighted outputs using expert evaluation. Furthermore, three distinct fine-tuning strategies are systematically investigated to enhance the performance of our model on the CHVR dataset. Experimental results indicate that ViT-HVE achieves the best performance in heritage value recognition tasks, and our proposed weighting method shows a high degree of consistency with expert evaluation results. The ViT-HVE evaluation method significantly enhances the objectivity and accuracy of cultural heritage recognition and evaluation, substantially reducing the cost and workload of related tasks, thereby offering scientific methodological support and practical guidance for the protection and utilization of cultural heritage.

While diffusion models have achieved remarkable progress in style transfer tasks, existing methods typically rely on fine-tuning or optimizing pre-trained models during inference, leading to high computational costs and challenges in balancing content preservation with style integration. To address these limitations, we introduce AttenST, a training-free attention-driven style transfer framework. Specifically, we propose a style-guided self-attention mechanism that conditions self-attention on the reference style by retaining the query of the content image while substituting its key and value with those from the style image, enabling effective style feature integration. To mitigate style information loss during inversion, we introduce a style-preserving inversion strategy that refines inversion accuracy through multiple resampling steps. Additionally, we propose a content-aware adaptive instance normalization, which integrates content statistics into the normalization process to optimize style fusion while mitigating the content degradation. Furthermore, we introduce a dual-feature cross-attention mechanism to fuse content and style features, ensuring a harmonious synthesis of structural fidelity and stylistic expression. Extensive experiments demonstrate that AttenST outperforms existing methods, achieving state-of-the-art performance in style transfer dataset.