OMNIMEDIA “Intelligent Rich Media  Search Platform”
Searchable Media for E-Learning As multimedia are being used to facilitate and enhance teaching and creative learning, making rich media including images and videos searchable are critical in today and future e-learning platform and management system.  Traditionally, it's a difficult challenge.  Some of the team members have invented an interdisciplinary framework which integrates the state-of-the-art image recognition technology and brain science research from MIT, intelligent cluster algorithm and patentable digital “tagging” technology.
Solution The solution is a community-based, video information-editing web platform, exposes and associates descriptive and searchable text with any digital video file.  It is diverging from the traditional closed-environment manual-labor model and the fully artificial intelligent approach to metadata content generation.  Instead, our solution is focusing on an automated solution to metadata generation through open, collaborative solutions.
Technology The technology is disruptive because a knowledge-sharing model for rich-media content management reduces maintenance costs and achieves higher production efficiency for metadata generation.  The platform simultaneously captures users and learners activities and preferences on the content provider’s own website.  It also facilitates incremental and collaborative learning through a open community.
OMNIMedia Innovation Artificial (Machine + Human) Collective Intelligence Media search Why? Search strategy: User-centered approach,  High accuracy More relevance (to user) User added business value Google gets it!  (PageRank algorithm leverages user contribution) Web Strategy: Ubiquitous
Why better? Accuracy Hand inserted metadata Cost effectiveness Machine learning / Computer Vision OMNIMedia  Platform
Features highlights Searchable video stream Non-linear video content access Crawlable by Search engine e.g. Google Dynamic in video advertisement (use image detection to detect sponsor images) Linkable - insert link (from one region of a frame in a clip to another frame in another clip) Unobtrusive - Text or bubble Ads to the clip
Example: MLB.com
  Object detection Cardeal.com Bookmark the  Info without Leaving the  video
Youtube Video NYC  Tiffany Bookmark the  Info without Leaving the  video
Object detection Great Sushi from Porter Sq, Cambridge Bookmark the  Info without Leaving the  video
Technology & Product
Image Recognition Input image Texture Based Classification Tree / Not-Tree Classification Feature Vector Decision Input Image Shape Based Classification car car ped Biological Model  Feature Extraction Car /  Not-Car Classification Feature Vector Decision
Image Recognition Performance Biological Model
Auto-tagging for photographs Landscape Group Shot Couple Portrait
Tagging assist
Clustering – add contextual info http://eigencluster.csail.mit.edu/
Interdisciplinary framework  (hybrid solution) OMNIMedia Cluster Algorithm OMNIMedia platform Auto-generate Initial  Metadata Add contextual  Info to further enhance metadata User enriched Media content Metadata – fine Tune the image Recognition  engine Computer Vision  (Image recognition)
Online Content Delivery Stack MySpace Operating System Browser ??? MLB.com YouTube etc G o o g l e WWW Internet
Online Content Delivery Stack Semantic Web / Metadata layer Internet OS (Browser and OS converge) Search www Internet G o o g l e OMNIMedia MySpace MLB.com YouTube etc ?
Technology Diffusion Web 2.0 services -> applications on InternetOS Data-driven G o o g l e OMNIMedia
Value Chain

Elearning rich media_search

  • 1.
    OMNIMEDIA “Intelligent RichMedia Search Platform”
  • 2.
    Searchable Media forE-Learning As multimedia are being used to facilitate and enhance teaching and creative learning, making rich media including images and videos searchable are critical in today and future e-learning platform and management system. Traditionally, it's a difficult challenge. Some of the team members have invented an interdisciplinary framework which integrates the state-of-the-art image recognition technology and brain science research from MIT, intelligent cluster algorithm and patentable digital “tagging” technology.
  • 3.
    Solution The solutionis a community-based, video information-editing web platform, exposes and associates descriptive and searchable text with any digital video file. It is diverging from the traditional closed-environment manual-labor model and the fully artificial intelligent approach to metadata content generation. Instead, our solution is focusing on an automated solution to metadata generation through open, collaborative solutions.
  • 4.
    Technology The technologyis disruptive because a knowledge-sharing model for rich-media content management reduces maintenance costs and achieves higher production efficiency for metadata generation. The platform simultaneously captures users and learners activities and preferences on the content provider’s own website. It also facilitates incremental and collaborative learning through a open community.
  • 5.
    OMNIMedia Innovation Artificial(Machine + Human) Collective Intelligence Media search Why? Search strategy: User-centered approach, High accuracy More relevance (to user) User added business value Google gets it! (PageRank algorithm leverages user contribution) Web Strategy: Ubiquitous
  • 6.
    Why better? AccuracyHand inserted metadata Cost effectiveness Machine learning / Computer Vision OMNIMedia Platform
  • 7.
    Features highlights Searchablevideo stream Non-linear video content access Crawlable by Search engine e.g. Google Dynamic in video advertisement (use image detection to detect sponsor images) Linkable - insert link (from one region of a frame in a clip to another frame in another clip) Unobtrusive - Text or bubble Ads to the clip
  • 8.
  • 9.
    Objectdetection Cardeal.com Bookmark the Info without Leaving the video
  • 10.
    Youtube Video NYC Tiffany Bookmark the Info without Leaving the video
  • 11.
    Object detection GreatSushi from Porter Sq, Cambridge Bookmark the Info without Leaving the video
  • 12.
  • 13.
    Image Recognition Inputimage Texture Based Classification Tree / Not-Tree Classification Feature Vector Decision Input Image Shape Based Classification car car ped Biological Model Feature Extraction Car / Not-Car Classification Feature Vector Decision
  • 14.
  • 15.
    Auto-tagging for photographsLandscape Group Shot Couple Portrait
  • 16.
  • 17.
    Clustering – addcontextual info http://eigencluster.csail.mit.edu/
  • 18.
    Interdisciplinary framework (hybrid solution) OMNIMedia Cluster Algorithm OMNIMedia platform Auto-generate Initial Metadata Add contextual Info to further enhance metadata User enriched Media content Metadata – fine Tune the image Recognition engine Computer Vision (Image recognition)
  • 19.
    Online Content DeliveryStack MySpace Operating System Browser ??? MLB.com YouTube etc G o o g l e WWW Internet
  • 20.
    Online Content DeliveryStack Semantic Web / Metadata layer Internet OS (Browser and OS converge) Search www Internet G o o g l e OMNIMedia MySpace MLB.com YouTube etc ?
  • 21.
    Technology Diffusion Web2.0 services -> applications on InternetOS Data-driven G o o g l e OMNIMedia
  • 22.

Editor's Notes

  • #8 Searchable video stream Non-linear video content access Crawlable - Video content -> text -> publish to web and crawled by search engine such as Google (and use Search Optimization technique) Dynamic - insert Ad (use image detection to detect sponsor images (e.g. ketchup) and dynamically insert link inside the video) Linkable - insert link (from one region of a frame in a clip to another frame in another clip – hypermedia) Unobtrusive - insert Text Ads to the clip on the side instead of a short clip at the opening