ICCV 2017 Tutorial Open Source Software and Datasets for Wildlife Video Surveillance October 22, 2017 http://www.viametoolkit.org/iccv-2017-tutorial/ Dr. Anthony Hoogs Matthew Dawkins Kitware, Inc. Dr. Charles Stewart, RPI Jason Holmberg, WildMe Dr. Concetto Spampinato Dr. Simone Palazzo University of Catania Dr. Ben Richards NOAA 1
Agenda Time Topic Presenter Introduction to Wildlife Surveillance 2:00 PM Datasets and OSS Anthony Hoogs, Kitware Underwater video datasets and the VIAME open-source framework for 2:15 PM fisheries stock assessment Anthony Hoogs, Kitware; Matt Dawkins, Kitware 3:15 PM break WildBook: Identifying and Re-identifying Individual Animals for the Purpose of Charles Stewart, RPI; 3:30 PM Population Assessment Jason Holmberg, WildMe Concetto Spampinato, University of Catania; Underwater Visual Data for Coral Reef Simone Palazzo, 4:30 PM Fish Biodiversity Monitoring University of Catania 5:30 PM Discussion and Summary All 2
Wildlife Visual Domains Terrestrial Aerial Underwater Courtesy of Tali Treibitz, Marine Imaging Lab, University of Haifa; The Morris Kahn Marine Research Station, University of Haifa 3 Computer Graphics and Multimedia lab, the Technion
Open Source Software https://github.com/wildbookorg Open source software for tracking individual animals in a population Data models and libraries Open source platform to integrate modules for fish detection, tracking and species recognition, classification, habitat classification Kitware Platforms ParaView CMake http://groups.inf.ed.ac.uk/f4k/ Open source software for fish detection, tracking and species recognition Ground truthing and annotation Resonant CDash 4
Fisheries Stock Assessment NOAA Fisheries Strategic Initiative on Automated Image Analysis Develop guidelines, set priorities, and fund projects to develop broad-scale, standardized, and efficient automated analysis of still and video imagery for use in stock assessment Video and Imagery Analytics for the Marine Environment http://viametoolkit.org/ https://github.com/kitware/viame Open source toolkit for fish and shellfish detection, tracking, classification, measurement Upcoming challenge: Fish detection and classification on underwater imagery 5
www.wildbook.org Wildbook blends structured wildlife research with artificial intelligence, citizen science, and computer vision based on identification of individual animals to speed population analysis and develop new insights to help fight extinction. Overview Origins of the Wildbook project A data management platform for biologists and citizen scientists A computer vision view of the problems Data sets and current results A way forward 6
Underwater Visual Data for Coral Reef Fish Biodiversity Monitoring Introduction to underwater monitoring Objectives and difficulties Approaches Technology Public datasets and initiatives FishCLEF and SeaCLEF challenges UniCT underwater background modeling datasets The Fish4Knowledge project code base Fish detection/tracking architecture Running the tool Extending the library 7
Underwater Visual Data for Fish Biodiversity Monitoring Video-based fish detection and tracking Image-based fish species recognition Image-based whale individual identification
Related Workshops Underwater Vision Workshop http://marine.acfr.usyd.edu.au/iccv13uv/ ICCV 2013 Automated Analysis of Video Data for Wildlife Surveillance http://marineresearchpartners.com/avdws2017/home.html WACV 2015, 2016, 2017 Visual Wildlife Monitoring https://www.iiitd.edu.in/~anands/iccvw2017_vwm/index.html ICCV 2017 Automated Analysis of Marine Video for Environmental Monitoring Proposal for CVPR 2018 9
Computer Vision and Environmental Monitoring Our community can help to address global issues environmental such as global warming, sustainability, endangered species protection Wildlife population changes Animal migrations Species collapse Wildlife trafficking Get involved! 10