M.W. Oram and D.I. Perrett. Modeling visual recognition from neurobiological constraints. Neural Networks, 7:945-972, 1994.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....grey level picture. PCA leads to a fast reduction of data by a linear transformation. We would like to remark, that from a biological point of view, in the human visual system there are no hints for data compression but a lot of hints for a data spreading in the rst stages of visual processing [31]. A problem of PCA methods is the restriction to linearity of transformations (for a discussion of this problem and some attempts to deal with it see [7] Within ORASSYLL non linear transformations (e.g. equation (5) and the criteria C1 and C2 in section 3) play an important role. In [17] it has ....

M.W. Oram and D.I. Perrett. Modeling visual recognition from neurobiological constraints. Neural Networks, 7:945-972, 1994.

....that for a robust pattern recognition a methodology is required which deals with uncertainties in a more intelligent manner especially at the higher levels of understanding. The anatomy of the primate form processing pathway indicates that the visual system employs a multi layered network [14]. The precise number of layers is debatable, but experiments have shown that normally there are seven or eight major cortical layers and at least four of them are employed for computation (see Fig. 1) The connections between the areas of primate visual form processing also suggest that four ....

....have been some demonstrative experiments that indicate that the sharp contours are more discriminate than gradients, straight lines are more discriminate than the irregular squiggles. It is known that the brain does not store the entire picture or patterns of the world but only some aspects of it [14]. Recognition of the objects is then based on these aspects and not on their hard copies. Studies of neural networks and neural science in general try to decipher the code of internal workings of brain [14] It has beendisclosed that some of the features are coded in the neurons, such as lines and ....

[Article contains additional citation context not shown here]

M.W. Oram and D.I. Perrett,"Modeling visual recognition from neurobiological constraints," Neural Networks, vol.7, No.6,7, pp.945-972, 1994.

....hard copies. Studies of neural networks and neural science in general try to decipher the code of internal workings of the brain [1] It has been disclosed that some of the features are coded in the neurons, such as lines and edges of various orientations, and junctions and end points of contours [16]. Contextual clues play a very important role in human handwriting recognition. It can be said that a human recognition system would produce much worse results in the absence of enough clues. The anatomy of the primate form processing pathway indicates that the visual system employs a ....

....clues play a very important role in human handwriting recognition. It can be said that a human recognition system would produce much worse results in the absence of enough clues. The anatomy of the primate form processing pathway indicates that the visual system employs a multi layered network [16]. The precise number of layers is debatable, but experiments have shown that normally there are seven or eight major cortical layers and at least four of them are employed for computation. The connections between the areas of primate visual form processing also suggest that four computational ....

[Article contains additional citation context not shown here]

M.W. Oram and D.I. Perrett, "Modeling visual recognition from neurobiological constraints," Neural Networks, vol.7, No.6,7, pp.945-972, 1994.

....grey level picture. PCA leads to a fast reduction of data by a linear transformation. We would like to remark, that from a biological point of view, in the human visual system there are no hints for data compression but a lot of hints for a data spreading in the first stages of visual processing [31]. A problem of PCA methods is the restriction to linearity of transformations (for a discussion of this problem and some attempts to deal with it see [7] Within ORASSYLL non linear transformations (e.g. equation (5) and the criteria C1 and C2 in section 3) play an important role. In [17] it ....

M.W. Oram and D.I. Perrett. Modeling visual recognition from neurobiological constraints. Neural Networks, 7:945--972, 1994.

.... about which a great deal is known (Hubel Wiesel, 1968; Szentagothai, 1977; Jones, 1981; Gilbert, 1983; Van Essen, 1985; Douglas Martin, 1998) 3) the response properties of neurons in each of the relevant brain areas have been well studied, and evolve from stage to stage in systematic ways (Oram Perrett, 1994; Logothetis Sheinberg, 1996; Tanaka, 1996) and (4) computer systems may already be powerful enough to emulate the functions of these neural processing stages, were it only known what exactly to do. A number of neuromorphic approaches to visual recognition have been proposed over the years ....

Oram, M. W., & Perrett, D. I. (1994). Modeling visual recognition from neurobiological constraints. Neural Networks, 7(6/7), 945--972.

.... and size) In terms of analogy to the processing in area V1 in the mammalian visual system C1 may be interpreted as the response of a certain column which indicates the general presence of a feature, whereas C2 represents the intercolumnar competition giving a more specific coding of this feature [11]. One shot learning: By positioning a rectangular grid on a roughly segmented object (see figure 6a,i) in front of homogeneous background and extracting significant features per instance as described above suitable representations of objects can already be extracted. These representations are ....

M.W. Oram and D.I. Perrett. Modeling visual recognition from neurobiological constraints. Neural Networks, 7:945--972, 1994.

....layers of quadratic SFA modules (SFA 2 ) with direct connectivity. It is clear from the architecture that the receptive eld size increases from bottom to top and therefore the units become potentially able to respond to more complex features, two properties characteristic for the visual system (Oram Perrett, 1994). 1a 1b 2b 2a 3a 3b 4a 4b SFA 1 SFA 2 SFA 2 SFA 1 SFA 1 SFA 2 SFA 1 SFA 2 1 9 9 17 17 33 65 65 33 retina Figure 1: A hierarchical network of SFA modules as a simple model of the visual system learning translation invariance. Di erent layers correspond to di erent cortical areas. For ....

Oram, M. W. and Perrett, D. I. (1994). Modeling visual recognition from neurobiological constraints. Neural Networks, 7(6/7):945-972.

....are described in section 4. We conclude with some general remarks related to the proposed method. 2. Multilevel fuzzy rule based pattern recognition The modeling of the visual recognition system from neurobiological constraints has shown that the visual system employs a multilayered network[13][15] The number of identified layers is dependent on the chosen model and is represented by a number of processing layers between two and seven[13] The same is valid for the designed fuzzy rule based pattern matching. It is a multifold recognition procedure with different levels of semantics. ....

.... The modeling of the visual recognition system from neurobiological constraints has shown that the visual system employs a multilayered network[13] 15] The number of identified layers is dependent on the chosen model and is represented by a number of processing layers between two and seven[13]. The same is valid for the designed fuzzy rule based pattern matching. It is a multifold recognition procedure with different levels of semantics. Let us assume that the pattern space is partitioned into various fuzzy subspaces. These subspaces represent the domains of the local fuzzy features. ....

[Article contains additional citation context not shown here]

M.W. Oram and D.I. Perrett,"Modeling visual recognition from neurobiological constraints," Neural Networks, vol.7, No.6,7, pp.945-972, 1994.

.... and sparse coding [6] There is also strong evidence that the human object recognition system processes the visual input through a couple of stages in which features of increasing complexity are extracted and in which Gabor wavelets represent only an early stage of processing (see, e.g. [7, 2]) Evaluating the second order statistics of Gabor wavelet responses we give statistical evidence for important second order relations for the class of natural images and therefore we support the understanding of the stages of the visual system beyond the extraction of Gabor wavelets responses. ....

M.W. Oram and D.I. Perrett. Modeling visual recognition from neurobiological constraints. Neural Networks, 7:945--972, 1994.

....(i.e. processed at very early stages of visual processing) arises from psychophysical experiments in (Treisman 1986) who showed that a curved line pops out in a set of straight lines. A question that is still open is the role of feedback in early stages of visual processing. It has been argued (Oram and Perrett 1994) that the short recognition time humans need for unknown objects (in the range of 100ms) makes computationally costly feedback loops unlikely. Others criticize this opinion, pointing to the huge amount of feedback connections between adjacent areas or to context sensitivity of cell responses (see, ....

....the visual cortex of primates hierarchical processing of features of increasing complexity and increasing receptive field size occurs. As a functional reason for processing of this type the advantages of capacity sharing, minimization of wiring length, and speed up have been mentioned (see e.g. (Oram and Perrett 1994)) Different coding schemes for features are discussed in the literature. The concept of local coding in which one neuron is responsible for one feature (Barlow 1972) leads to problems: Because for each possible feature a separate neuron has to be used, a large amount of neurons is required. ....

[Article contains additional citation context not shown here]

Oram, M. and Perrett, D. 1994. Modeling visual recognition from neurobiological constraints. Neural Networks, 7:945--972.

....and Computer Engineering University of Queensland St. Lucia Q. 4072. Australia. fstratton, tdg elec.uq. edu.au ABSTRACT: Cortical neurons tuned to specific stimuli, such as orientation selective cells of area V1, have been found to respond with greater vigour when the stimulus is unexpected ([1]) Other neurons have been found which become active in anticipation of a stimulus which has not yet arrived. This paper introduces a neural architecture and accompanying unsupervised learning algorithm which can account for these observed characteristics. Computations that can be performed by ....

....these observed characteristics. Computations that can be performed by this architecture are suggested, and simulations show how it can be applied to problems of image completion and novelty detection. Keywords: Neural networks, expectation, unsupervised learning 1 Introduction Oram and Perret ([1]) presented evidence which showed that about half of the neurons in monkey cortex that they tested in the visual and tactile modalities were more responsive when an input to which they were tuned was not expected than when it was. When a monkey had control of the position of a stimulus and it ....

Mike W. Oram and David I. Perret, "Modeling visual recognition from neurobiological constraints", Neural Networks, vol. 7, no. 6/7, pp. 945--972, 1994.

.... about which a great deal is known (Hubel Wiesel, 1968; Szentagothai, 1977; Jones, 1981; Gilbert, 1983; Van Essen, 1985; Douglas Martin, 1998) 3) the response properties of neurons in each of the relevant brain areas have been examined, and evolve from stage to stage in systematic ways (Oram Perrett, 1994; Logothetis Sheinberg, 1996; Tanaka, 1996) and (4) present day computer systems may already be powerful enough to emulate the functions of these neural processing stages, were it only known what exactly to do. A number of neuromorphic approaches to visual recognition have been proposed over ....

Oram, M. W., & Perrett, D. I. (1994). Modeling Visual Recognition from Neurobiological Constraints. Neural Networks, 7 (6/7), 945-972.

....(i.e. processed at very early stages of visual processing) arises from psychophysical experiments in (Treisman, 1986) who showed that a curved line pops out in a set of straight lines. A question that is still open is the role of feedback in early stages of visual processing. It has been argued (Oram and Perrett, 1994) that the short recognition time humans need for unknown objects (in the range of 100ms) makes computationally costly feedback loops unlikely. Others criticize this opinion, pointing to the huge amount of feedback connections between adjacent areas or to context sensitivity of cell responses (see, ....

....the visual cortex of primates hierarchical processing of features of increasing complexity and increasing receptive field size occurs. As a functional reason for processing of this type the advantages of capacity sharing, minimization of wiring length, and speed up have been mentioned (see e.g. (Oram and Perrett, 1994)) Different coding schemes for features are discussed in the literature. The concept of local coding , in which one neuron is responsible for one feature (Barlow, 1972) leads to problems: Because for each possible feature a separate neuron has to be used, a large amount of neurons is required. ....

[Article contains additional citation context not shown here]

Oram, M. and Perrett, D. (1994). Modeling visual recognition from neurobiological constraints. Neural Networks, 7:945--972.

....representation combined with our hierarchical feature processing allows a fast and effective locating (section 4. using EGM. Our system has certain analogies to the visual system of vertebrates. There is evidence for curvature sensitive features processed in a hierarchical manner in early stages [3]; sparse coding is discussed as a coding scheme used in the visual system [1] and metric organization of features seems to play an important role for information processing in the brain [3] We aim to apply these concepts in our artificial object recognition system. 2. The Banana Space The ....

.... There is evidence for curvature sensitive features processed in a hierarchical manner in early stages [3] sparse coding is discussed as a coding scheme used in the visual system [1] and metric organization of features seems to play an important role for information processing in the brain [3]. We aim to apply these concepts in our artificial object recognition system. 2. The Banana Space The principle P3 gives us a significant reduction of the search space. Instead of allowing, e.g. all linear filters as possible features, we restrict ourself to a small subset. Considering the risk ....

[Article contains additional citation context not shown here]

M.W. Oram and David I. Perrett, "Modeling Visual Recognition from Neurobiological Constraints," Neural Networks, vol. 7, pp. 945--972, 1994.

No context found.

M.W. Oram and D.I. Perrett. Modeling visual recognition from neurobiological constraints. Neural Networks, 7:945-972, 1994.

No context found.

M.W. Oram and D.I. Perrett, "Modeling visual recognition from neurobiological constraints," Neural Networks, Vol.7, No.6-7, pp. 945-972, 1994.

No context found.

M. W. Oram and D. I. Perret. Modeling visual recognition from neurobiological constraints. Neural Networks, 7(6-7):945--972, 1994.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC