Obtaining annotations for complex computer vision tasks such as object
detection is an expensive and time-intense endeavor involving a large number of
human workers or expert opinions. Reducing the amount of annotations required
while maintaining algorithm performance is, therefore, de