Relationships in instance segmentation and anomaly detection

In the field of computer vision, the overarching goal is to achieve a human level of image understanding through a computational system. Progress in the field has fueled discoveries in broader artificial intelligence and machine learning, generalizing to other signals (e.g., text, sound) and goals (e.g., winning games, understanding language). Because achieving human-level perception is currently unachievable with general artificial intelligence, the field’s progress is largely measured within specific tasks. Many major advances have been motivated by these well-defined tasks, and the results have in turn driven advancements in more general artifical intelligence and machine learning.