Year: 2025
Type: article
Abstract: In the ceramic packaging industry, there is an ever-growing and escalating demand for unique and culturally resonant visual effects. However, traditional prediction methods encounter difficulties when attempting to seamlessly blend multimodal data sources like images, text, and profound cultural insights. This frequently results in inaccurate visual effect forecasts and may even cause potential cultural misinterpretations. To surmount these constraints, this paper introduces the Visual Multimodal Inference and Synthesis for Intelligent Ceramic Packaging (VISIC). It constructs a hierarchical multimodal feature fusion network, refines the Light-GAN, and incorporates a cultural compliance verification module. Specifically, the model employs advanced algorithms to more effectively manage data. Experiments demonstrate that VISIC improves multimodal feature extraction accuracy by 5.08% and attains a peak prediction success rate of 82.6%, significantly enhancing the prediction capabilities for ceramic packaging visual effects.
Authors: Zhou Long, Junzhe Ouyang
Institution: Jingdezhen Ceramic Institute
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W4411670898