1. Visual Cognition II Object Perception

2. Theories of Object Recognition Template matching models Feature matching Models Recognition-by-components Configural models

3. Template matching Detect patterns by matching visual input with a set of templates stored in memory – see if any template matches. TEST INSTANCE “J” TEMPLATE “T” TEMPLATE

4. Problem: what if the object differs slightly from the template? E.g., it is rotated or scaled differently? Solution: use a set of transformations to best align the object with a template (using translation, rotation, scaling) TEST INSTANCE “J” TEMPLATE “T” TEMPLATE rotation

5. Template-matching works well in constrained environments

6. Figure 2-15 (p. 58) Examples of the letter M. Problem: template matching is not powerful enough for general object recognition

7. Feature Theories Detect objects by the presence of features Each object is broken down into features E.g. A = + +

8. Problem Many objects consist of the same collection of features Need to also know how the features relate to each other  structural theories One theory is recognition by components Different objects, similar sets of features

9. Recognition by Components Biederman (1987): Complex objects are made up of arrangements of basic, component parts: geons. “Alphabet” of 24 geons Recognition involves recognizing object elements (geons) and their configuration

11. Why these geons? Choice of shape vocabulary seems a bit arbitrary However, choice of geons was based on non-accidental properties. The same geon can be recognized across a variety of different perspectives: except for a few “accidental” views:

12. Viewpoint Invariance

13. Viewpoint invariance is possible except for a few accidental viewpoints, where geons cannot be uniquely identified

14. Prediction Recognition is easier when geons can be recovered Disrupting vertices disrupts geon processing more than just deleting parts of lines Object Deleting line segments Deleting vertices

15. Evidence from priming experiments

16. Problem Theory does not say how color, texture and small details are processed. These are often important to tell apart similar objects. E.g.:

17. Configural models of recognition Individual instances are not stored; what is stored is an “exemplar” or representative element of a category Recognition based on “distance” between perceived item and prototype prototype match “Face space” no match

18. Configural Models

19. Configural effects in face processing Do these faces have anything in common? How about these ones? By disrupting holistic processing, it becomes easier to process the individual parts

20. Face superiority effect Farah (1994)

21. Results

22. Face superiority effect Parts of faces are not processed independently. The context of other face parts (e.g. mouth) influences recognition of a particular part (e.g. nose) Face superiority effect disappears when face is inverted

23. Context and Object Recognition

24. What do you see?

25. Top-down vs. Bottom up Visual Input Low Level Vision High Level Vision Bottom-up processing  Stimulus driven Knowledge

26. Top-down vs. Bottom up Visual Input Low Level Vision High Level Vision Top-down processing  Knowledge driven  Context Effects Knowledge

27. Top-down effects: knowledge influences perception

28. Slide from Rob Goldstone

29. Problem for many object recognition theories. How to model role of context? Context can often help in identification of an object Later identification of objects is more accurate when object is embedded in coherent context

30. Context can alter the interpretation of an object

31. Context Effects in Letter Perception The word superiority effect: discriminating between letters is easier in the context of a word than as letters alone or in the context of a nonword string. DEMO: http://psiexp.ss.uci.edu/research/teachingP140C/demos/demo_wordsuperiorityeffect.ppt http://psiexp.ss.uci.edu/research/teachingP140C/demos/demo_wordsuperiorityeffect.ppt (Reicher, 1969)

32. Word superiority effect suggests that information at the word level might affect interpretation at the letter level Interactive activation model: neural network model for how different information processing levels interact Levels interact –bottom up: how letters combine to form words –top-down: how words affect detectability of letters

33. The Interactive Activation Model Three levels: feature, letter, and word level Nodes represent features, letters and words; each has an activation level Connections between nodes are excitatory or inhibitory Activation flows from feature to letter to word level and back to letter level (McClelland & Rumelhart, 1981)

34. The Interactive Activation Model PDP: parallel distributed processing Bottom-up: –feature to word level Top-down: –word back to letter level Model predicts Word superiority effect because of top-down processing (McClelland & Rumelhart, 1981)

35. Predictions of the IA model – stimulus is “WORK” At word level, evidence for “WORK” accumulates over time Small initial increase for “WORD” WORK WORD WEAR

36. Predictions of the IA model – stimulus is “WORK” At letter level, evidence for “K” accumulates over time – boost from word level “D” is never activated because of inhibitory influence from feature level K R D

37. For a demo of the IA model, see: http://www.itee.uq.edu.au/~cogs2010/cmc/chapters/LetterPerception/

38. Take-home message What you see is not what is out there in the outside world (ie., not like “taking a picture”), but instead a result of visual computation -- only those computations that are critical for survival, shaped by the evolution.