DALES: Automated Tool for Detection, Annotation, Labelling and Segmentation of Multiple Objects in Multi-Camera Video Streams

摘要

In this paper, we propose a new software tool called DALES to extract semantic information from multi-view videos based on the analysis of their visual content. Our system is fully automatic and is well suited for multi-camera environment. Once the multi-view video sequences are loaded into DALES, our software performs the detection, counting, and segmentation of the visual objects evolving in the provided video streams. Then, these objects of interest are processed in order to be labelled, and the related frames are thus annotated with the corresponding semantic content. Moreover, a textual script is automatically generated with the video annotations. DALES system shows excellent performance in terms of accuracy and computational speed and is robustly designed to ensure view synchronization.