Deep learning hasn't (yet) rendered classical computer vision obsolete. Why some challenges are still best solved with traditional algorithms.

Then, it uses that knowledge to create new, artificial image-mask pairs to augment a small dataset of real examples. A segmentation model is trained using both.

Over the past decades, computer scientists have developed many computational tools that can analyze and interpret images.