In this report we consider the problem of rapidly annotating a video with bounding boxes for a novel object. We describe a UI and associated workflow designed to make this process fast for an arbitrary novel target.
生成讲解博客的中文版本
生成AI智能校验的中文版本