Over the past few years, Generative AI has emerged as an innovative tool by generating high-quality images and videos based on the development of deep learning technology. New architectures such as GAN, VAE, and Diffusion models have enabled creative and practical applications in various industries such as entertainment, advertising, and scientific simulations. However, the generation model is difficult to evaluate because it produces results without correct answers, thus various evaluation scales are being studied to solve this problem. This presentation demonstrates how to apply the evaluation metrics for various tasks including in-painting, image generation through black box optimization, and so on.
I hope this lecture will give you a lot of inspiration on how to evaluate and use the generated image.