Informedia Video Information Summarization & Demonstration Testbed

About VACE I |  Ongoing VACE I Project Info  |  VACE I Reports  ||  VACE II  |  Sponsor

Informedia VACE I: Module Integration

Based on the discussion in the ARDA VACE Workshop (June 20-21, 2001) at Carnegie Mellon University, Informedia is going to provide the following two items to facilitate the integration effort for the VACE program.

1. Metadata Exchange Interface: It uses extensible markup language (XML) to specify a module's functionality, input data, and detected objects with their attributes.

   Viewing note: A better way to view the XML and DTD file is to view the source page directly (using "View Source" or "Page Source"). Web browser will probably NOT be able to display an XML's DTD file. However, the DTD page is indeed loaded while an error message is being displayed. To view the DTD file, just use "View Source".

   Syntax: The document type definition (DTD) of the Metadata Exchange Interface XML is defined in

   • version 1.1 revision 2 (two revisions by HNC): "vace_module_v1_1_r2.dtd"
   • version 1.2 (event description added by USC): "vace_module_v1_2.dtd"

   Semantics: The semantics of all elements and attributes are documented mainly as the attribute values in a module template XML file

   • verion 1.1 revision 2: "vace_module_v1_1_r2.xml"
   • verion 1.2: "vace_module_v1_2.xml"

   Two examples using XML for Metadata Exchange Interface are provided:

   • Object detection example: face recognition
   • Object tracking example: vehicle tracking
     
     

2. Informedia Video Data: Informedia is providing an hour broadcast news video as part of the testbed for the VACE program.
   • Broadcast news video (531 MB)

   In addition, 732 keyframes, which are identified by Informedia's color-histogram-based scene-break detection, are being provided:

   • in JPEG format (5.1 MB, .tar.gz file)
   • in uncompressed (PPM) format (121.3 MB, .tar.gz file)

   The Metadata Exchange Interface (input portion) for this video is also available:

   • Without keyframe specification (1 KB)
   • With keyframe specification (216 KB)

Other Related Information

UMD VACE Video Collection

The LAMP Lab at University of Maryland, College Park provides a collection of sample videos, including those given by ARDA.

Boeing VACE Video Collection

Boeing Space and Communications provides the VACE community two videos in conjunction with Boeing's work on generating super-resolution images:
• booktitles1.avi (metadata)
• booktitles2.avi (metadata)