I’m trying to solve an anomaly detection task using an autoencoder model, but I’m getting poor results after training. My teacher suggested I use data augmentation to batch on each step during epoch. I’m not sure how to do this.

I’ve searched online and found information about using ImageDataGenerator and augmentation layers. However, I think these methods don’t apply augmentations to batch on each step during epoch.

Thank you for help and sorry if my question is not really clear.