* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Download PowerPoint Presentation - VARS Overview
Survey
Document related concepts
Transcript
MBARI’s Video Annotation and Reference System (VARS) Brian Schlining and Nancy Jacobsen-Stout MBARI MBARI MBARI Video Data • More than 300 ROV dives per year • More than 12,000 hours of video observations generated in 16 years • More than 1.6 million individual observations Scientific Applications Video observations and related data • Used by MBARI researchers in well over 200 professional publications and presentations, in addition to countless education and outreach projects • Span multiple sub-disciplines within biology, geology, chemistry “A picture can be worth a thousand words” (if you can find it!) Benthic Diversity Interesting Taxa (gorgonocephalus) Annotations • Trained science staff identify animal taxonomy, animal behavior, geological features, evidence of human impact, and other objects recorded in the ROV video stream • Annotations are stored in a database shared between MBARI scientists. • Database continues to evolve as our researchers learn more • Work closely with and rely on scientists to keep updated on new discoveries and publications in a timely fashion Annotation History 1st generation – Free Text – Inconsistent annotations – Difficult to search 2nd generation – VICKI/VIMS – Constrained annotations – Written in Smalltalk – Generated files 3rd generation – VARS – Constrained annotations – Written in Java – Writes information directly to database VARS - System Components Knowledge Base = hierarchical constraint lexicon & reference More than 3500 taxonomic, geologic, technical concepts. Reference media and descriptions. Annotation = catalog video observations, inc. visual media Uses the KB data to create and edit structured annotations for objects in the video stream. Query = retrieve video-related data Uses the KB to structure simple to complex queries against annotations database VARS - Knowledge Base Manages constraint lexicon and references VARS - Annotation Generates video-annotations and frame-grabs Frame capture (QuickTime for Java) VCR Control (RS-422) VARS - Annotation Generates video-annotations and frame-grabs Knowledge Base Customizable Interface VARS - Query Searches for and retrieves information and frame-grabs. VARS - Query Searches for and retrieves information and frame-grabs. VARS - Query Searches for and retrieves information and frame-grabs. Example – “Cold Seeps” Vesicomyid Clams Vestimentiferan Tubeworms “Cold Seep” Map Vintage 1999 “Cold Seep” Map Vintage 2004 Monterey bay is one of the most observed continental margins in the world Visited 0.9% pixels in this map Example – “Cold Seeps” • Previous “seep” models assumed seeps are associated with faults (fluid low conduits) Chemosynthetic Biological Communities occur preferentially on steep slopes? Steep slopes imply areas of recent erosion. VARS Development Goals VARS user goals include: 1) Provide responsive user interfaces to maximize user efficiency and accommodate real- and greater than real-time analyses. 2) Provide a robust system. 3) Provide clear and intuitive user interfaces. VARS developmental goals include: 1) Create maintainable (modular, understandable) software for future developers. 2) Maximize use of existing software (e.g. old VIMS query and commercially available software) 3) Provide a solution that can be exported to other institutions. 4) The database should be simple to maintain by MBARI IS administrators. VARS Requirements Vars system requirements: 1) Java 1.4 or greater 2) Java Comm (or RXTX) 3) Quicktime for Java (For frame-capture) 4) Relational Database VARS Deployment VARS Deployment Point Lobos Shore Western Flyer VARS Deployment Point Lobos Shore Western Flyer VARS - Annotation data VARS - Annotation data A related group of tapes. e.g. tapes from a single expedition VARS - Annotation data A video source, such as a tape or QuickTime movie VARS - Annotation data A single video frame VARS - Annotation data An object observed in a particular video-frame. The name is constrained by the knowledgebase VARS - Annotation data Descriptive information. Can also be used to link ‘concepts’. For example, ‘nanomia eating krill’ VARS - Annotation data Position, physical data, expedition information, and camera data, VARS - Annotation data VARS - Knowledgebase data VARS - Knowledgebase data ~ Hierarchical, Allows representation of relations or phylogeny ~ Linked to descriptive info. (History, authors, images, video, etc. [not shown]) VARS - Knowledgebase data ~ A concept may have one or more concept-names. ~ One concept-name is designated as a primary name. ~ User may annotate with any concept-name. However, only the primary name is stored in the database. VARS - Knowledgebase data VARS – Available to you VARS is available to other science institutions. See http://www.mbari.org/vars/ Fall 2005 – Release VARS as open-source December 2005 – MBARI will host a developer workshop If you are interested in using VARS contact: Judith Connor ([email protected]) Director of Information & Technology Dissemination