A Survey on Semi-Automated and Automated Approaches for Video Annotation

Video analytics systems have recently gained intensive attention due to the fact that they play a practical role in a broad range of topics, including understanding scenes in autonomous driving and processing huge volumes of video data. Nevertheless, labeling frame-by-frame of video datasets is extremely cost-prohibiting and time-consuming. Therefore, the development of effective annotation systems will be crucial for the generation of proper annotations for large-scale datasets of video. This survey paper briefly reviews novel representative literature and systematizes the commonalities in different machine learning techniques such as supervised learning, active learning, transfer learning, and neural networks. We consider some reviews that introduce annotation tools as well. Ultimately, we summarized the performances of all the mentioned approaches applied to various datasets and demonstrated the experimental results.