AI to Judge AI-Generated Images: Both Semantics and Perception Matter
Jean Atsumi Flaherty, Chibuike Onuoha, Incheon Paik, Truong Cong Thang · 2023
Recent advances in AI-generated contents (AIGC) could revolutionize the digital content creation in many applications, such as e-learning and communication. The goal of this study is to predict the perceptual quality of AI-generated images. While the existing studies just focus on perceptual features only, our hypothesis is that both perceptual features and semantic features are important. Through experiments, we show that combining these two kind of features can improve prediction performance. To our best knowledge, this is the first work that applies multiple AI tools to predict this kind of quality.