Self-supervised representation learning anomaly detection methodology based on boosting algorithms enhanced by data augmentation using StyleGAN for manufacturing imbalanced data.