How to use data augmentation for object detection in tensorflow 2

Question

I\'m planning to rewrite my old TF1/Keras data generation to use the tf.data and tf.image APIs. In particular, I would like use the augmentation method