CoVidA: pen-based collaborative video annotation

Abstract

In this paper, we propose a pen-based annotation tool for videos. Annotating videos is an exhausting task, but it has a great benefit for several communities, as labeled ground truth data is the foundation for supervised machine learning approaches. Thus, there is need for an easy-to-use tool which assists users with labeling even complex structures. For outlining and labeling the shape of an object, we introduce a pen-based interface which combines pen and touch input. In our experiments we show that especially for complex structures the usage of a pen device improves the effectiveness of the outlining process.