TAPe+ML v2: How we make a single recognition model for any computer vision task (opens in new tab)

In computer vision systems, a separate stack is assembled for each task: one backbone for classification, another for detection, and a…