Automated image generation

• Main Author: Chen, Xueyao
• Other Authors: Lu Shijian
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2023
• Subjects: Engineering::Computer science and engineering
• Online Access: https://hdl.handle.net/10356/166132
• Institution: Nanyang Technological University

Recent years have shown emerging trends in the role of artificial intelligence in art and image generation, as the popularity of AI image generators such as DALL·E, MidJourney, Stable Diffusion and NovelAI have proven the many benefits and uses of automated image generation. The goal of this project is to explore the technology behind AI image generation and create a prototype of an automated image generator service. Generative adversarial networks were used in the project to create an image generator capable of generating new images according to predefined parameters (colour, subject). The images used for training are public domain images. The results were collected by conducting a survey with ten participants. The participants were asked to classify the generated images as “real” and “fake”. The results were then compiled to determine the overall accuracy of the generated images.