Utilization of Generative Artificial Intelligence in Visual Effects
Yan Du · 2024
The research on computer vision mainly aims to enable computers to understand and interpret images. Computer vision is an emerging discipline, and applying generative artificial intelligence technology to image processing is an important entry point in current machine vision research. When generating images, this article utilized adversarial learning mechanisms to optimize the parameters of the generator and discriminator, creating a competitive relationship between the two and gradually improving the quality of the generated images. Adversarial training methods were used in video processing to optimize the generated model, enabling it to map between different styles and achieve style transfer. The generative adversarial network was optimized using a cyclic consistency loss function to ensure consistency of video content before and after editing. Based on the trained recurrent generative adversarial network, style transformation and content editing on the video were performed. The PSNR value of generated image 3 was 27.1; the PSNR (peak signal-to-noise ratio) value of the original image was 27.8; the SSIM (Structural Similarity) value of the generated image was 0.95; the SSIM value of the original image was 1.00. The PSNR and SSIM values of the generated image with image number 3 in the experiment were relatively high and close to the corresponding values of the original image. This article provided a reference for the research on the application of generative artificial intelligence in visual effects.