Semi-supervised Visual Tracking of Marine Animals Using Autonomous Underwater Vehicles

In-situ visual observations of marine organisms is crucial to developing behavioural understandings and their relations surrounding ecosystem. Typically, these are collected via divers, tags, remotely-operated or human-piloted vehicles. Recently, however, autonomous underwater vehicles equipped with cameras embedded computers GPU capabilities being developed for a variety applications, in particular, can be used supplement existing data collection mechanisms where human operation tags more difficult. Existing approaches have focused on using fully-supervised tracking methods, but labelled many species severely lacking. Semi-supervised trackers may offer alternative solutions because they require less than counterparts. However, there not realistic datasets, the performance semi-supervised algorithms domain well understood. To better evaluate utility, this paper we provide (1) novel dataset specific animals located at http://warp.whoi.edu/vmat/, (2) an evaluation state-of-the-art context animal tracking, (3) real-world through demonstrations algorithm on-board vehicle track wild.