Augmenting images training data to add variability doesn't always end well since the augmentation changes the semantic content of the image. Here's how to solve it.
A review and clear explanation of the NeRF method, which can be used to synthesize 3D scenes out of an input image. This method is the base for many other research methods that followed.