Abstract
With the large-scale development of urban landscaping construction, the problems in its expression of beauty are gradually emerging. This article selects the beauty of ceramic art in landscape design, which combines complexity and ambiguity, as the research object and conducts a systematic study on its aesthetic evaluation. Ceramic artists use pattern content to express their attitude towards life and pursuit of art. But ceramic art creation is difficult and highly professional, making it difficult for non-professionals to get started. Against the backdrop of the rapid development of artificial intelligence, image analysis technology has emerged, allowing non-professionals to decorate ceramic products. In this study, the extraction, quantification, and recognition of ceramic features are carried out through machines. The idea and implementation method of replacing experts with machines for intelligent ceramic recognition are explored. Traditional models such as Convolutional Neural Network (CNN), Visual Geometry Group (VGG) network, and fast migration model are introduced, and the improved fast migration model are proposed. Finally, through experimental comparison, the results show that the improved fast migration model exhibits robust changes in extraction quality as the learning amount increases, and has generalization ability. It is an effective new method for achieving ceramic aesthetic research and analysis, and has good practicality.
Introduction
Ceramic art is a cornerstone of traditional landscape design, offering unique textures, colors, and cultural depth. However, integrating it effectively requires high-level professional expertise. For non-professionals, the "eye recognition" of quality ceramic patterns is subjective and difficult to quantify. As urban landscaping expands, there is a growing gap between the demand for beautiful, culturally rich environments and the availability of experts to design them. This research explores replacing the "expert eye" with a machine learning model capable of extracting, quantifying, and recognizing ceramic aesthetic features.
Methodology: The Improved Fast Migration Model
The research builds upon established computer vision architectures but introduces a specific refinement for aesthetic "migration" (style transfer). The study utilizes VGG19, a deep convolutional neural network known for its effectiveness in feature extraction. By removing the fully connected layers and focusing on the convolutional layers, the model acts as a "loss network" to measure perceptual and semantic differences between images.
Unlike original style transfer methods that require thousands of iterations for a single image, the proposed feedforward network generates results in a single pass. The core novelty lies in the training strategy: instead of using a single fixed image as a style target, the model is trained sequentially on multiple images of the same style, generating checkpoints ckpt1, ckpt2, and ckpt3.
This iterative optimization allows the model to capture the "essence" of a style rather than just the quirks of a single reference image, leading to better generalization. The model balances content loss (preserving landscape structure) and type/style loss (applying ceramic aesthetics) using Euclidean distance and Gram matrices.
Experimental Results
The study utilized a specialized dataset of 1,046 ceramic images (732 for training, 209 for validation, and 105 for testing) sourced from museum collections. This dataset focuses on traditional patterns like blue and white porcelain, including figures, animals, plants, and landscapes.
Experimental results recorded using MATLAB show a clear hierarchy of performance. The Improved Fast Migration Model reached an accuracy (P value) of 94.87%, significantly outperforming the standard fast migration model (82.43%) and basic CNN (77.01%).
Furthermore, the improved model showed a 20.90% increase in recall rate over the traditional CNN. Qualitative analysis via Structural Similarity (SSIM) confirmed that the improved model produces images much closer to the "real effect" of ceramic art, successfully preserving core features like characters and traditional instruments while applying the desired artistic style.
Conclusion
The proposed model democratizes ceramic art application in landscape design by allowing non-professionals to achieve professional-grade aesthetic consistency. By moving to a feedforward "fast migration" model, analysis time is reduced from hours to seconds. While the model excels at capturing general style, future work will address challenges in fine line details and background complexity, potentially by incorporating image segmentation structures to apply different levels of migration to foreground and background elements.
Summary updated