Research on the Application of Artificial Intelligence Algorithms in Image Content Creation in the Environment of Digital Transformation of Film Industry

Abstract

This paper proposes a new architecture for stylizing 3D meshes based on textual information based on the promotion of artificial intelligence technology for the digital transformation of the film industry. Combined with the object beam analysis of bi-directional reflection distribution function, the CLIP model is applied to the parameter optimization of a given frame to realize the 3D mesh stylization for realism rendering. FID, LPIPS and CLIP score metrics are used to analyze the performance of the model generated based on CLIP model-driven 3D mesh stylization. And the image content generated by the model is scored. Among them, the image content generated by this method scores 36.85 and 39.14 in accuracy and authenticity, respectively.The similarity value between text and image tested on MS-COCO dataset reaches 0.312, and the result measured by this paper with VQA evaluation metrics on CelebA-HQ dataset is 0.894.Combining the performances of all the metrics, this paper proposes the CLIP-based model-driven 3D network stylized generation model can be used for movie image content creation and can be further promoted.