PSYCH 240 1st Edition Lecture 4Outline of Last Lecture I. Processes of VisionOutline of Current Lecture I. Pattern RecognitionII. Template TheoryIII. Feature TheoryIV. Structural DescriptionsV. Next Monday’s LectureCurrent Lecture: Bottom Up ProcessingI. Pattern Recognitiona. Representationi. The proximal image projected on our retina which is then delivered as a mental representation to our brainii. This does NOT recognize what the image it isb. Matchingi. Recognizing the imageii. Taking the percept and matching it against our memoryc. Shape Constancyi. The proximal stimulus is projected differently but the actual object is still the sameii. The proximal stimulus looks different depending on the view of the object yet we can still recognize the objectII. Template Theory– a template against which the percept is being matched (think of children’s shape puzzles)a. Applicationsi. Perceptual representation1. In order to recognize an object, we compare it to images of the similar objects that we have stored in memoryii. Memory representation1. We compare the object to a variety of stored images and identify it by the image it most closely resemblesb. Problemsi. Transformations 1. Objects are viewed from different viewpoints => delivering different proximal stimulia. A cursive ‘E’ looks different from a print ‘E’ii. Obstructed objects1. A butterfly half behind a leaf is still a butterfly but if a template theory was used all that can be seen is half of an image and it would not be recognized as a butterflyiii. Visual system decompositionThese notes represent a detailed interpretation of the professor’s lecture. GradeBuddy is best used as a supplement to your own notes, not as a substitute.1. Decomposes scenes into primitive features – doesn’t fit images to their own individual templateIII. Feature Theory– Our perceptual system delivers a type of representation that is stored in memory a. Applicationsi. Perceptual representation1. Uppercase letter “E”a. 3 horizontal linesb. 1 vertical linec. 4 right anglesii. Memory Representation1. Uppercase letter “E” vs. uppercase “F”a. Ei. 3 horizontal linesii. 1 vertical lineiii. 4 right anglesb. Fi. 2 horizontal linesii. 1 vertical lineiii. 3 right anglesb. Evidencei. Physiology (recordings from neurons)1. On-off areas of our neuronsii. Stabilized retinal images1. Our eyeballs are constantly shaking so that our photoreceptors don’t get worn out a. if our eyeballs didn’t shake our rods and cones would have to keep firing to the point that they wouldn’t stopb. the image we see is stationary iii. Pandemonium model (computer program)1. First is uses feature demons to look for specific physical features in the image and each detector tells us how many of each feature is present2. Then it uses cognitive demons that recognize specific sets of features (patterns)3. Next, specific cognitive demons scream that the image is a “b” or an “r” (if the image is aletter) 4. The decision demon listens to the loudest shout in pandemonium to identify the correct letteriv. Visual search1. People are better at recognizing caricatures than actual photographsa. Makes sense because we recognize people based on specific features and a caricature emphasizes those features2. Searching for specific letters in jumbled listc. Problemsi. Lack of spatial relationships between features1. Lower case d vs. b vs pa. All are 1 line and 1 half circleii. Different arrangements of the same features produce different objects1. Think mug vs. bucket or briefcase vs. drawera. They have the same features but they are rearranged spaciallyiii. Recognition by componentsiv. Non-accidental propertiesIV. Structural Descriptionsa. Recognition-by-Components (RBC): Matching Processi. Created by Irv Biedermanii. The objects we perceive in the world can be represented by an alphabet of geometric forms (geons)1. The brick geon, the cylinder geon, the piece of pie geoniii. Steps in the Process1. Detect elementary features, edges2. Find non-accidental propertiesa. Some features in the visual world are an accident of our viewpoint but others can only be non-accidental (they always look this way)i. straight or curvedii. parallel or nonparalleliii. co-termination or no co-terminationiv. symmetrical or asymmetrical3. Determine component geona. Brick here, a cone here, and another brick here – this is how we match the instrument4. Match to memorya. Cylinder, cylinder with a brick on top = flashlightiv. Evidence1. Partial or degraded objectsa. Deleting accidental and non-accidental propertiesi. A picture that is partially obscured or not drawn in (disconnected lines)ii. Easier to decipher when accidental properties but when non-accidental properties are deleted people have trouble deciphering the image2. Object Complexitya. The more complicated objects are easier to recognize when in geon form i. It is easier to pick out a more complicated mix of geons because there are fewer objects that fit the descriptive image3. Unusual Orientationsa. If you lose the non-accidental properties because of an unusual orientation , it makes it significantly harder to identify the objecti. Identifying a blender just by looking at a picture that is angled directly above it (only see the top of the blender)v. Problems1. Similar objectsa. Human faces – all humans have the same geons for their noses, mouths, eyes, heads, etc.i. Yet, we have no trouble deciphering the subtle differences between faces2. Lack of brain evidence3. Context effects (next lecture)*all of the theories are purely bottom-up- none of these theories use any top-down processing in their theories but we’ll cover those next weekvi. Applications (things it is good at deciphering1. Transformationsa. Uses non-accidental properties that don’t depend on viewpoint (how far away the object is, whether it is rotated or scaled, etc.)2. Relationships between featuresa. Build geons then figure out the spatial relationships among those geons so even if the same geons are present, it may be different objects3. Explaining how we make sense of nonsense objectsa. We can make sense of objects we have never seen beforeV. Next Lecturea. Top-down effectsb. Read 107-121 in the