This VOT2020 workshop is the eighth in a series of annual workshops on visual single-object tracking of a priori unknown objects. The five challenges hosted by the workshop are the VOT-ST on RGB short-term tracking, the VOT-RT on RGB short-term real-time tracking, the VOT-LT on RGB long-term tracking, the VOT-RGBT on RGB and thermal tracking and the VOT-RGBD on RGB and depth long-term tracking. A significant change has been made to VOT-ST by replacing the bounding box by a segmentation mask as label. Our aim is to promote trackers capable of per-pixel object localization and to close the gap between the areas of single-object tracking and video segmentation. The evaluation methodology of VOT-ST has been adapted for modern trackers by introducing a new failure recovery. The adapted toolkit was re-implemented in Python and makes the MATLAB version obsolete. The first paper covers the challenges' results, comprising work of 111 coauthors from 46 institutions. The paper (with results omitted) was shared with all co-authors who provided feedback and improved the quality. The VOT2020 Organizing Committee (OC) received five regular paper submissions which were reviewed in a double-blind process; each paper by three independent Program Committee (PC) members. We received additionally a rejected ECCV paper which included all reviews, rebuttal reports, and information about improvements to the original work. This paper was reviewed by the OC. All six papers were finally accepted to the workshop.
展开▼
