To promote the inheritance of traditional culture, a variety of emerging methods rooted in machine learning and deep learning have been introduced. Dunhuang patterns, an important part of traditional Chinese culture, are difficult to collect in large numbers due to their limited availability. However, existing text-to-image methods are computationally intensive and struggle to capture fine details and complex semantic relationships in text and images. To address these challenges, this paper proposes the Diffusion Adapter Network (DANet). It employs a lightweight adapter module to extract visual structural information, enabling the diffusion model to generate Dunhuang patterns with high accuracy, while eliminating the need for expensive fine-tuning of the original model. The attention adapter incorporates a multihead attention module (MHAM) to enhance image modality cues, allowing the model to focus more effectively on key information. A multiscale attention module (MSAM) is employed to capture features at different scales, thereby providing more precise generative guidance. In addition, an adaptive control mechanism (ACM) dynamically adjusts the guidance coefficients across feature layers to further enhance generation quality. In addition, incorporating a cross-entropy loss function enhances the model’s capability in semantic understanding and the classification of Dunhuang patterns. The DANet achieves state-of-the-art (SOTA) performance on the proposed Diversified Dunhuang Patterns Dataset (DDHP). Specifically, it attains a perceptual similarity score (LPIPS) of 0.498, a graph matching score (CLIP score) of 0.533, and a feature similarity score (CLIP-I) of 0.772.

Understanding how the built environments of historic districts shape tourist experiences and behavioral intentions is important for their sustainable development. Drawing on scene theory and the Stimulus–Organism–Response (SOR) model, this study employs structural equation modeling (SEM) and fuzzy-set qualitative comparative analysis (fsQCA) to empirically analyze 498 valid questionnaire responses collected in Pingjiang Historic District, Suzhou, China. The study examines the relationships among built environmental stimuli, psychological experience, and tourist behavioral intentions. The findings show that: (1) Tourist behavioral intentions are associated with multiple interacting conditions rather than a single linear factor; (2) A distinctive cultural atmosphere and sustained cultural knowledge experiences may provide conditions that support tourists’ attachment to the district and are associated with intentions to revisit, recommend, and share their experiences. Rich sensory experiences and locally distinctive activities may contribute to hedonic experience, place attachment, and cultural identity, thereby supporting positive behavioral intentions. In addition, integrating cultural knowledge into interactions among tourists and between tourists and local groups may provide a context in which hedonic experience and positive behavioral intentions are more likely to emerge. (3) Among the psychological experience variables, strengthening tourists’ identification with local culture can provide a relatively stable psychological basis for revisiting, recommending, and sharing historic-district experiences.

With the rise of AI-generated content in creative fields like film and animation, assessing its visual perception quality has become a major challenge. Traditional evaluation methods often fall short. This study presents ZCASNet, a new hybrid model for automated evaluation of AI-generated short films based on several significant indicators like visual coherence, scene rationality, detail reality, temporal consistency. In ZCASNet framework, Convolutional Neural Networks (CNN), ResNet50, is applied to extract spatial features from video frames. Multi-Head Attention (MHA) captures semantic and temporal relationships across frames. The Zebra Optimization Algorithm (ZOA) enhances the CNN-MHA architecture by refining hyperparameters to improve its performance. Subsequently, a Support Vector Machine (SVM) is employed to classify the optimized features. Our experiments show that ZCASNet outperforms baseline models, including CNN-only, CNN with MHA, Transformer models, and standard SVM. ZCASNet achieves an accuracy of 92.4%, demonstrating notable performance. Ablation studies confirm the importance of the RBF kernel function and each module in the model architecture. Overall, ZCASNet offers an effective and interpretable solution for evaluation in AI-generated video. It has potential uses in content recommendation, AI-assisted filmmaking, and quality control in multimedia applications.