Smarter document extraction starts here.

Abstract: To address the problem of long-tail distribution for the large vocabulary object detection task, existing methods usually divide the whole categories into several groups and treat each group ...