accuracy and look at every letter on page, but this will take more time and the cost of cognitive capacity needed is huge to read every single letter -When reading, we are reading some of the letters and making inferences about what’s not filled in for most part. These inferences would be safe bet in reality because most of time is correct lMaking inferences about words: eg. “Yuo cna porbalby raed tihs esaliy desptie teh msispeillgns” This can be read correctly because the brain is making inferences to fill in things not known by using known words lThe feature net model with letter and bigram detectors is the classic model applied only to words, but it can be applied to objects as well Recognition by components model (RBC model): lThis model helps explain object recognition which includes additional detectors called Geons or geometric icons lThese are basic building blocks of objects which are shown in the image lIrving Biederman determines that we only need three dozen geons to describe every object in the world, similar to use 26 letters in all possible combinations to form every word in English language lJust like feature detectors, RBC model uses low-level and high-level detectors. Feature detectors respond to edges, curves which activates the geon detectors. Higher level responds to combinations of geons lActivating geon assembly represents relation between geons eg. Five geons can create different combinations as 6 objects we recognize lAccording to this model, this allows us to recognize different features at different angles. This model is known to be viewpoint independentwhich means no matter what position is relative to an object, people will be able to recognize it Viewpoint dependent model: lAnother model for recognition of objects is that we have stored different
viewpoints of each object in memory. lBut people can store so many views eg. For the car in the image, many views such as from side, front, top and back etc. lThis model suggests that recognition sometimes requires rotation i.e. recognition is slower at some viewpoint eg. Store the view of car on the side. In order to recognize it from other positions, in our mind, it has to be mentally rotated which delays recognition of the same object lThis is known as viewpoint dependent model which the viewpoints of objects influence recognition lIn this model, there is a hierarchy of detectors and each layer is concerned with more complex aspects. With detectors to represent each of the viewpoint of object, some detectors fire more strongly for particular views. This is how we achieve object recognition in this model lFor recognition of objects, there is a network type model with hierarchical detectors in place which helps us recognize objects in the world Face recognition: lThere is lots of evidence that faces are special in a cognitive point of view. For example, there is a special neural area dedicated to processing faces differently from other things which is known as fusiform face area l
You've reached the end of your free preview.
Want to read all 48 pages?
- Winter '08
- temporal lobe, John B. Watson, The Man Who Mistook His Wife for a Hat