Semi-automatic image and video annotation system for generating ground truth information

Abstract

Recently, techniques for automatically interpreting images or videos through machine learning based on big data have been actively studied. In this paper, we propose a semiautomatic image and video annotation system to generate ground truth information, which is essential information for machine learning of images or videos. Unlike the conventional methods for generating simple ground truth information manually, the proposed system not only provides various ground truth information such as object information, motion information, and event information, but also uses a semi-automatic image and video annotation method for fast generation of ground truth information. The ground truth information generated by the proposed system is stored in the metadata database as a form of XML. The implementation results show that the proposed system provides not only fast ground truth annotation, but also more various ground truth information compared to the existing methods.