1 / 37

Perceiving and Representing Structured Information using Objects

Perceiving and Representing Structured Information using Objects. Organization. Image based theories Structural 3D theories Object perception: Object displays. Template theories based on 2D image processing. In structural theories we extract the structure of a scene in terms of 3D primitives.

dfannie
Download Presentation

Perceiving and Representing Structured Information using Objects

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Perceiving and Representing Structured Informationusing Objects

  2. Organization • Image based theories • Structural 3D theories • Object perception: Object displays

  3. Template theories based on 2D image processing In structural theories we extract the structure of a scene in terms of 3D primitives Image Based vs Structure Theories

  4. Template theories A template with simple morphing operations

  5. Objects as patterns of (2D) patterns

  6. Scenes (gist in 100 msec) Patterns of patterns Torralba and Oliva

  7. Properties of Image Recognition • Remarkable image recognition memory • Up to 5 images for second in object identification • Applications in image searching interfaces • RSVP Rapid serial visual presentation

  8. Perception of object structure

  9. Geon Theory (Irving Biederman) 3D Primitives “Geons” Structural skeleton Shape from shading is also primitive

  10. How Does the Brain Find Geons?

  11. Canonical Views To make things clear both component parts and the critical joints must be clear

  12. Processing stages for recognition

  13. The Geon Diagram (Irani) • 3D shape primitives for architecture - entities and relationships • Surface texture and color for attributes

  14. Pattern finding& Recognition 13% errors: 4.3 sec sub-structure 22% memory errors 26% errors 7.1 sec sub-structure 42% memory errors

  15. On top of - relies on Underneath – support, foundation Inside – containment (private code) Attachment points (external interfaces; part_of relationships) more than topology mid Upper Lower Semantics

  16. Natural semantics Instances Dependency Multiplicity Strength of Relationship

  17. The object file: Triesman and Kaheman

  18. 2 ½ D design • Use 3D objects to represent entities • Layout to make structure clear in 2D. • -Connections, basic shapes • Use canonical views

  19. The Object Display (Wickens) • Use complex objects to fuse variables • Map entities to object parts • Map structure to object structure • Can be pictorial (simplified) – an engine + fuel tank • Map attributes to object attributes - color, size, motion etc.

  20. Vol : 45 L Temp: 45 deg Flow: 20% Solids: 23% Object Display 100 L 45 L 0.2/1.0 45 deg Solids: 23%

  21. Approx 1/3 errors with anesthesiologists Dr George Blike

  22. A medical ventilator • Such displays • Must be custom

  23. Object displays may lead to lower errors Lend themselves to overly literal interpretation. They are not general purpose. Object displays must be custom designed More abstract representations leave more room for interpretation (better support the creative process) The concrete/abstraction tradeoff

  24. Chart Junk (ducks) Tufteor Effective Communication 1990 2000 2010 1990 2000 2010

  25. A Duck? (Tufte)

  26. Holding attention-Object scene conflict Used in advertising

  27. Emotional and motivational effects

  28. Conclusion • There is a great future in 3D object icons and object structures • It is happening anyway, but nice to know that it is not always “chart junk”.

  29. Conclusions • Visual space is non-homogenous • The plane orthogonal to the line of sight and the forward direction are very different. • A 21/2D design attitude recognizes the cost of actions.

More Related