UIBAIFED - Artificial Intelligence Facial Expression Dataset

UIBAIFED, a novel facial expression dataset designed to enhance Facial Expression Recognition (FER) by providing high-quality, realistic images labeled with detailed demographic attributes, including age group, gender, and ethnicity. Unlike existing datasets, UIBAIFED incorporates a fine-grained classification of 22 micro-expressions, based on the universal facial expressions defined by Ekman and the micro-expression taxonomy proposed by Gary Faigin. The dataset was generated using Stable Diffusion and validated through a convolutional neural network (CNN), achieving an accuracy of 82% in expression classification. The results highlight the dataset’s reliability and potential to improve FER systems. UIBAIFED fills a critical gap in the field by offering a more comprehensive labeling system, enabling future research on expression recognition across different demographic groups and advancing the robustness of FER models in diverse applications.