Title:	SEMI-AUTOMATIC MULTI-CAMERA VIDEO ANNOTATION TOOL FOR OBJECT TRACKING AND OPTICAL FLOW BENCHMARK CREATION
Author(s):	Vladimir V. Devyatkov and Igor I. Lychkov
ISBN:	978-989-8533-91-3
Editors:	Katherine Blashki and Yingcai Xiao
Year:	2019
Edition:	Single
Keywords:	Moving Object Tracking, Multi-Camera Video, Optical Flow, Video Annotation Tool
Type:	Short Paper
First Page:	427
Last Page:	431
Language:	English
Cover:
Full Contents:	click to dowload
Paper Abstract:	This paper addresses a problem of video annotation for object tracking evaluation. Contemporary object tracking methods relies heavily on optical flow calculation techniques. State of the art optical flow calculation methods are not enough to provide robust and accurate tracking under object occlusion and shading conditions. Thus, comprehensive visual object tracking datasets are required for deep analysis, benchmarking and improvement of object tracking methods. In most cases publicly available real video datasets does not include ground truth for optical flow evaluation, so new datasets might be in demand. Public video annotation tools for object tracking dataset labeling were reviewed. Lack of open source video annotation tools with functions of automatic optical flow markers and object bounding box visual tracking was noticed. In this work we developed a new open video annotation tool that is suitable for object bounding box labeling as well as optical flow annotation in multicamera video recordings. This tool makes it possible to specify the feature point and/or moving object position on the first video frame and track its position automatically in the subsequent video frames. This tool supports manual adjustment of the tracked position if needed. Multi-layer modular architecture of the annotation tool simplifies its testing, maintenance and further extension by the community.
	Go Back