Synthetic images of fear and disgust facial expression generated for dataset skewness reduction in Facial Expression Recognition

These images were generated through the Hugging Face Serverless Inference API, which unfortunately is not available anymore, and has been refactored into the Inference Providers API. Text-to-image models were used to generate images of disgust and fear facial expressions, to increase the instances of those two classes , which are the two least-represented in the AffectNet dataset.