GAIA: A Benchmark of Analyzing User Rankings for Synthesized Images

Text-to-image models which are a part of Generative AI have become essential tools for digital artists and enthusiasts to create visually captivating images. These models have garnered significant attention and rapid advancements in recent years, enabling the creation of realistic and visually appealing images from textual descriptions. However, assessing the quality of these generated images remains a challenging task due to varying perceptions of image quality. Additionally, generated images often lack clear ground truth and the intricate details that capture human attention. To address these challenges in the study of artificially generated images, we introduce a novel approach with the Generative Artificial Image Assessment (GAIA) dataset. This dataset includes images from eight popular text-to-image AI models along with user rankings. GAIA is evaluated and predicted by pre-trained state-of-the-art networks using ranking classes and a regression technique to analyze the images. Our approach combines objective evaluation metrics, subjective human judgment, benchmark datasets with diverse ground truth annotations, and advancements in multimodal learning techniques. This comprehensive methodology provides a pathway to advancing the field of text-to-image generation.