S. Lawrence, C. L. Giles, A. C. Tsoi and A. D. Back, "Face Recognition: A Convolutional Neural Networks Approach," IEEE Trans. on Neural Networks, Special Issue on Neural Networks and Pattern Recognition, Vol. 8, No. 1, pp. 98-113, 1997.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....to some degree characteristics of the human visual cortex and encompass scale and translation invariant feature detection layers. Convolutional neural networks have been successfully applied for character recognition [5] object detection [5] and more speci cally for the task of face recognition [3]. 2 Convolutional Neural Networks Figure 1 shows the architecture of the convolutional neural networks we trained for the task of facial expression recognition. Its layers alternate between convolution layers with feature maps ConvLay ConvLay k;l = g(I k;l W k;l B k;l ) 1) and ....

....to increase the learning speed, we normalized also the variances of the input variables by dividing them by their standard deviation in of the images of the training set: I norm = Iin Iin in . No attempts were taken to reduce image dimensionality by using e.g. holistic PCA as demonstrated in [3]. Instead, we relied on the kernels of the feature extraction layers to perform decorrelation of the input data. Holistically applied PCA without using sophisticated pose normalization procedures would attempt to represent pose information, which is not desired, as there are too many pose ....

Steve Lawrence, C. Lee Giles, A.C. Tsoi, and A.D. Back. Face recognition: A convolutional neural network approach. IEEE Transactions on Neural Networks, 8(1):98-113, 1997.

.... necessary if the feature extraction method is robust against varying illumination) There are many approaches to face recognition ranging from the Principal Component Analysis (PCA) approach (also known as eigenfaces) 40, 63] Elastic Graph Matching (EGM) 14, 33] Artificial Neural Networks [34, 64], to pseudo 2D Hidden Markov Models (HMM) 17, 53] All these systems differ in terms of the feature extraction procedure and or the classification technique used. These systems, and many others, are described in the sections below. 2.1 Geometric Features vs Templates Brunelli and Poggio [5] ....

....increased. To speed up the search, a randomly IDIAP RR 03 20 selected subset of pixels is used instead of the entire image. Verification results on a database which had mainly expression changes show a minor improvement over Duc s extended EGM approach (described in Section 2. 4) Lawrence et al. [34] proposed the use of a hybrid neural network approach to face recognition. The system combined local image sampling, a self organizing map (SOM) 30] and a convolutional neural network. On a database of 40 people, the proposed approach obtained an identification error rate of 3.8 , compared to ....

S. Lawrence, C. L. Giles, A. C. Tsoi and A. D. Back, "Face Recognition: A Convolutional Neural-Network Approach", IEEE Trans. Neural Networks, Vol. 8, No. 1, 1997, pp. 98-113.

....maps [6, 7, 2] Self organizing feature maps have previously been used in computer vision, for example, in image compression [1] medical image processing [11] Email: betkeoak.bc.edu, http: oak.bc.edu betke. The author has been supported by NSF grant EIA 9871219. and face recognition [8] applications. In com puter vision, the term feature generally refers to a local image property, for example, a line or circle in the edge map of an image. As discussed in Ref. 4] edge based methods may discard a significant amount of information pertinent to an object s recognition and so may ....

S. Lawrence, C. Giles, A. Tsoi, A. Back, "Face Recognition: A Convolutional NeuralNetwork Approach," IEEE Trans. on Neural Networks, Vol. 8, No. 1, pp. 98-111, 1997.

.... and the results of a generic algorithm (GA) weighted classifier shows a 96 recognition rate. The use of neural networks in the classification of complex images, for instance, human faces, is limited to less than 20 classes. There are exceptions to such a limited approach: Lawrence et al. [12] have developed a multiple neural network system comprising a Kohonen self organising feature map (SOM) 11] for quantizing a set of input image samples into a topological space, and a backpropagation network, also called a convolution network, that learns to incorporate constraints that allow it ....

Lawrence, S., Giles, C.L., Ah Chung Tsoi & Back, A.D. (1997). Face Recognition: A Convolutional Neural Network Approach. IEEE Transactions on Neural Networks, vol. 8 (1) pp 98-113.

.... necessary if the feature extraction method is robust against varying illumination) There are many approaches to face recognition ranging from the Principal Component Analysis (PCA) approach (also known as eigenfaces) 39, 61] Elastic Graph Matching (EGM) 13, 32] Artificial Neural Networks [33, 62], to pseudo 2D Hidden Markov Models (HMM) 16, 52] All these systems differ in terms of the feature extraction procedure and or the classification technique used. These systems, and many others, are described in the sections below. 2.1 Geometric Features vs Templates Brunelli and Poggio [5] ....

....increased. To speed up the search, a randomly IDIAP RR 03 20 selected subset of pixels is used instead of the entire image. Verification results on a database which had mainly expression changes show a minor improvement over Duc s extended EGM approach (described in Section 2. 4) Lawrence et al. [33] proposed the use of a hybrid neural network approach to face recognition. The system combined local image sampling, a self organizing map (SOM) 29] and a convolutional neural network. On a database of 40 people, the proposed approach obtained an identification error rate of 3.8 , compared to ....

S. Lawrence, C. L. Giles, A. C. Tsoi and A. D. Back, "Face Recognition: A Convolutional Neural-Network Approach", IEEE Trans. Neural Networks, Vol. 8, No. 1, 1997, pp. 98-113. IDIAP--RR 03-20

.... necessary if the feature extraction method is robust against varying illumination) There are many approaches to face recognition ranging from the Principal Component Analysis (PCA) approach (also known as eigenfaces) 86, 150] Elastic Graph Matching (EGM) 34, 74] Artificial Neural Networks [73, 151], to pseudo 2D Hidden Markov Models (HMM) 36, 123] All these systems di#er in terms of the feature extraction procedure and or the classification technique used. These systems, and many others, are described in the sections below. It must be noted that while the verification task has the ....

....score is increased. To speed up the search, a randomly selected subset of pixels is used instead of the entire image. Verification results on a database which had mainly expression changes show a minor improvement over Duc s extended EGM approach (described in Section 5.2. 4) Lawrence et al. [73] proposed the use of a hybrid neural network approach to face recognition. The system combined local image sampling, a self organizing map (SOM) 70] and a convolutional neural network. On a database of 40 people, the proposed approach obtained an identification error rate of 3.8 , compared to ....

S. Lawrence, C. L. Giles, A. C. Tsoi and A. D. Back, "Face Recognition: A Convolutional Neural-Network Approach", IEEE Trans. Neural Networks, Vol. 8, No. 1, 1997, pp. 98-113.

....compute the correlation between a face and one or more model templates to estimate the face identity. Statistical tools such as Support Vector Machines (SVM) 30, 20] Linear Discriminant Analysis (LDA) 2] Principal Component Analysis (PCA) 28, 29] Kernel Methods [26, 16] and Neural Networks [25, 11, 15] have been used to construct a suitable set of face templates. While these templates can be viewed as features, they mostly capture global features of the face images. Facial occlusion is often difficult to handle in these approaches. The geometry feature based methods analyze explicit local ....

Steve Lawrence, C. Lee Giles, Ah Chung Tsoi, and Andrew D. Back. Face recognition: A convolutional neural network approach. IEEE Transactions on Neural Networks, 8(1):98--113, 1998.

....to some degree characteristics of the human visual cortex and encompass scale and translation invariant feature detection layers. Convolutional neural networks have been successfully applied for character recognition [5] object detection [5] and more specifically for the task of face recognition [3]. 2 Convolutional Neural Networks Figure 1 shows the architecture of the convolutional neural networks we trained for the task of facial expression recognition. Its layers alternate between convolution layers with feature maps 99 9 60 ....

....the learning speed, we normalized also the variances of the input variables by dividing them by their standard deviation the images of the training set: 0870 . No attempts were taken to reduce image dimensionality by using e.g. holistic PCA as demonstrated in [3]. Instead, we relied on the kernels of the feature extraction layers to perform decorrelation of the input data. Holistically applied PCA without using sophisticated pose normalization procedures would attempt to represent pose information, which is not desired, as there are too many pose ....

S. Lawrence, C. L. Giles, A. Tsoi, and A. Back. Face recognition: A convolutional neural network approach. IEEE Transactions on Neural Networks, 8(1):98--113, 1997.

....features like local edges into more complex patterns like corners and T junctions. Methods which were proposed so far for the optimization of these features include unsupervised competitive learning combined with manually designed training patterns [2] supervised gradient based optimization [4], enumeration heuristics [7] and sparse coding [13] Few works use evolutionary methods to optimize hierarchical vision systems. Pan et al. 6] optimize features with manually designed patterns as targets in intermediate stages of the vision architecture. Shi et al. 12] on the contrary use ....

....the fitness of a single individual. The question is how strongly generalization is affected by this problem. Therefore, it is important to check the generalization ability of the final result again with a validation dataset. For this test we use the COIL100 [5] data base and the ORL test dataset [4], containing face images. We first performed an optimization of the nonlinearity parameters alone, using a feature set of 50 combination features, that were obtained according to a local combination enumeration as suggested by [7] For this feature set, after manual tuning of the nonlinearities ....

[Article contains additional citation context not shown here]

S. Lawrence, C. L. Giles, A. C. Tsoi and A. D. Back. Face recognition: A convolutional neural-network. IEEE Trans. Neur. Netw., 8(1):98-- 113, 1997.

....we used the ORL face image dataset (copyright AT ; T Research Labs, Cambridge) which contains 10 images each of 40 people with variability in expression and pose. Without any parameter or feature modification we obtain a classification rate of 96 using 5 training views, compared to 96.5 [6] using gradient based supervised learning on higher hierarchical stages. Another central ability for visual recognition is the rejection of unknown stimuli. With an identical setting as described above, however, using a single sigmoidal output VTU with s = tanh(r 2) we performed gradient based ....

S. Lawrence, C. L. Giles, A. C. Tsoi, and A.D. Back. Face recognition: A convolutional neural-network approach. IEEE Trans. Neut. Netw., 8(1):98-113, 1997.

.... necessary if the feature extraction method is robust against varying illumination) There are many approaches to face recognition ranging from the Principal Component Analysis (PCA) approach (also known as eigenfaces) 83, 145] Elastic Graph Matching (EGM) 33, 70] Artificial Neural Networks [69], to pseudo 2D Hidden Markov Models (HMM) 35, 119] All these systems differ in terms of the feature extraction procedure and or the classification technique used. These systems, and many others, are described in the sections below. It must be noted that while the verification task has the ....

....score is increased. To speed up the search, a randomly selected subset of pixels is used instead of the entire image. Verification results on a database which had mainly expression changes show a minor improvement over Duc s extended EGM approach (described in Section 5.2. 4) Lawrence et al. [69] proposed the use of a hybrid neural network approach to face recognition. The system combined local image sampling, a self organizing map (SOM) 66] and a convolutional neural network. On a database of 40 people, the proposed approach obtained an identification error rate of 3.8 , compared to ....

S. Lawrence, C. L. Giles, A. C. Tsoi and A.D. Back, "Face Recognition: A Convolutional Neural-Network Approach", IEEE Trans. Neural Networks, Vol. 8, No. 1, 1997, pp. 98-113.

....equally well in the verification scenario. There are many approaches to face based systems ranging from the ubiquitous Principal Component Analysis (PCA) approach (also known as eigenfaces) 2] Dynamic Link Architecture (also known as elastic graph matching) 3] Artificial Neural Networks [4], to pseudo 2D Hidden Markov Models (HMM) 5] These systems differ in terms of the feature extraction procedure and or the classification technique used. For example, in [2] PCA is used for feature extraction and a nearest neighbour classifier is utilized for recognition. In [3] biologically ....

S. Lawrence et al., "Face Recognition: A Convolutional NeuralNetwork Approach", IEEE Trans. Neural Net., Vol. 8, No. 1, 1997.

....smaller the size of a role is, the easier it is to interpret. 1.2. Neural Network Ensembles When these problems have existed within the context of the traditional neural network models such as the Multi Layer Perceptron (MLP) several solutions have been proposed using neural network ensembles [15, 13, 1]. These solutions normally take the form of a collection of single neural networks connected in some fashion that work together to produce more robust classifiers. To generate a single output class, another layer of the classification system is added that finds a consensus between the individual ....

S. Lawrence, C. L. Giles, A. Tsoi, and A. Back. Face recognition: A convolutional neural network approach. IEEE Transactions on Neural Networks, 8(1):98-113, 1997.

....tested on the ORL dataset. The results in this table are also plotted in g. 4 for better clarity. Clearly, the proposed approach outperforms those based on Volumetric Frequency Domain Representation [11] and Standard Hidden Markov Models [12] Convolutional and Probabilistic Neural Networks [13, 14] are also improved up to a high level of con dence. Only Nearest Feature Line [15] Support Vector Machines [16] and Embedded Hidden Markov Models [17] give similar results. Our approach, however, is much simpler and still maintains a clear advantage in terms of error rate. Table 1. Error rates ....

....approaches tested on the ORL dataset. Approach Error 95 conf. rate interval Local Features based Nearest Neighbour 0 0.0 1.8 Embedded Hidden Markov Models [17] 2.0 0.6 5.0 Support Vector Machines [16] 3.0 1.1 6.4 Nearest Feature Line [15] 3.0 1.1 6. 4 Convolutional Neural Networks [13] 4.0 1.7 7.7 Probabilistic Neural Networks [14] 4.0 1.7 7.7 Standard Hidden Markov Models [12] 7.5 4.3 12.1 Volumetric Frequency Domain Rep. 11] 7.5 4.3 12.1 4 6 8 12 14 LFNN EHMM SVM NFL CNN PNN SHMM VFD Fig. 4. Error rates and their corresponding 95 con dence interval ....

A. Lawrence, C. Giles, A. Tsoi, and A. Back. Face recognition: A convolutional neural network approach. IEEE Trans. on Neural Networks, 8(1):98-113, 1997.

....95 con dence interval with those obtained by other successful approaches tested on the ORL dataset. Clearly, the proposed approach outperforms those based on Volumetric Frequency Domain Representation [7] and Standard Hidden Markov Models [8] Convolutional and Probabilistic Neural Networks [9, 10] are also improved up to a high level of con dence. Only Nearest Feature Line [11] Support Vector Machines [12] and Embedded Hidden Markov Models [13] give similar results. Our approach, however, is much simpler and still maintains a clear advantage in terms of error rate. 4 Conclusions A ....

....tested on the ORL dataset. Approach Error 95 conf. rate interval Local Features based Nearest Neighbour 0.5 0.0 2.8 Embedded Hidden Markov Models [13] 2.0 0.6 5.0 Support Vector Machines [12] 3.0 1.1 6.4 Nearest Feature Line [11] 3.0 1.1 6. 4 Convolutional Neural Networks [9] 4.0 1.7 7.7 Probabilistic Neural Networks [10] 4.0 1.7 7.7 Standard Hidden Markov Models [8] 7.5 4.3 12.1 Volumetric Frequency Domain Rep. 7] 7.5 4.3 12.1 ....

A. Lawrence, C. Giles, A. Tsoi, and A. Back. Face recognition: A convolutional neural network approach. IEEE Trans. on Neural Networks, 8(1):98-113, 1997.

....successfully form the core components of complex human imaging systems. The security system developed by Lin et al. [9] proved that neural nets can compete with other state of the art core imaging techniques; combinations of various types of neural nets can also form complete recognition systems [7]. However, neural nets are still most commonly viewed as specialist components for classification. The SOM has previously been applied to imaging with varying degrees of success. Perhaps the most notable work is the PicSOM generic image retrieval system [6] a number of tree structured SOMs are ....

S. Lawrence, C. L. Giles, A. C. Tsoi, and A. D. Black. Face Recognition: A Convolutional Neural Network Approach. IEEE Transactions on Neural Networks, 8(1):98-- 113, 1997.

....the correlation between a face and one or more model templates to estimate the face identity. Statistical tools such as Support Vector Machines (SVM) 30, 21] Linear Discriminant Analysis (LDA) 2] Principal Component Analysis (PCA) 27, 29, 11] Kernel Methods [25, 17] and Neural Networks [24, 7, 12, 16] have been used to construct a suitable set of face templates. While these templates can be viewed as features, they mostly capture global features of the face images. Facial occlusion is often difficult to handle in these approaches. The geometry feature based methods analyze explicit local ....

Steve Lawrence, C. Lee Giles, Ah Chung Tsoi, and Andrew D. Back. Face recognition: A convolutional neural network approach. IEEE Transactions on Neural Networks, 8(1):98--113, 1998.

....The Eigenfaces Algorithm In the eigenfaces method, faces are projected into eigen space by a Karhunen Lo eve transform [ 24, 7 ] avoiding reliance on semantic knowledge. 1 Since identi cation appears to depend on good intensity correlation between images, substantial pre processing is needed [ 19 ] Changes in scale are a principal cause of error for this method if no correction were to be made [ 36 ] Allowance for lighting and contrast variation [ 22 ] and ane transformation of the head position are also made. Changes in facial expression and time variation in facial appearance 1 ....

S. Lawrence, C. L. Giles, and A. D. Tsoi, A. C.and Back. Face recognition: A convolutional neural-network approach. IEEE Transactions on Neural Networks, 8(1):98-114, January 1997.

....smaller the size of a rule is, the easier it is to interpret. 1.2. Neural Network Ensembles When these problems have existed within the context of the traditional neural network models such as the Multi Layer Perceptron (MLP) several solutions have been proposed using neural network ensembles [15, 13, 1]. These solutions normally take the form of a collection of single neural networks connected in some fashion that work together to produce more robust classifiers. To generate a single output class, another layer of the classification system is added that finds a consensus between the individual ....

S. Lawrence, C. L. Giles, A. Tsoi, and A. Back. Face recognition: A convolutional neural network approach. IEEE Transactions on Neural Networks, 8(1):98--113, 1997.

....and normalised. Work on face analysis was performed as long ago as the 1970s [13] with key work twenty years later on PCA techniques by Kirby and Sirovich [14] and Turk and Pentland [15] Subsequent studies have applied a wide variety of approaches, including various types of neural networks [16, 17], Hidden Markov Models (HMMs) 18] and shape analysis [19] The components of a fully automatic face recognition system include: face detection to determine whether or not a face is present [19] face location and segmentation to determine the boundaries of the face [20] signal processing or ....

S. Lawrence, C. Giles, A. Tsoi, and A. Back. Face Recognition: A Convolutional Neural Network Approach. IEEE Trans. Neural Networks, pages 98--113, 1997.

.... Convolutional Networks have been applied to many applications, among others: handwriting recognition ( LeCun et al. 1989) Martin 1993) as well as machine printed character recognition ( Wang and Jean 1993) on line handwriting recognition ( Bengio et al. 1995) and face recognition ((Lawrence et al. 1997)) Fixed size convolutional networks that share weights along a single temporal dimension are known as Time Delay Neural Networks (TDNNs) and applied widely in speech processing and time series prediction. Variable size convolutional networks, which have applications in object detection and ....

Lawrence, S., Giles, C. L., Tsoi, A. C., and Back, A. D. (1997). Face Recognition: A Convolutional Neural Network Approach. IEEE Transactions on Neural Networks, 8(1):98--113.

....of training samples, the training samples were randomly shuffled after every training loop. For the ORL database, the number of outputs of the MLP was always 40 and a winner take all strategy was used for classification. To allow comparisons, the same training and test set sizes are used as in [7, 9] , i.e. the first 5 images for each subject are the training images and the remaining 5 images are used for testing. Hence there are 200 training images and 200 test images in total and no overlap exists between the training and test images. Due to the small size of the available data, a ....

.... using the MATLAB benchmark utility and the published SPEC approach recognition rate training time recognition time relative best mean speed HMM [9] 87 eigenfaces(PCA) 9] 90 P2D HMM [9] 95 4 minutes y 1 192 convolutional NN [7] 98.5 z 96.2 x 0.004 x 4 hours 0:5 seconds 1 MLP [8] 84.0 77.2 0.0353 10 minutes k 0.0014 seconds k 89 DCT MLP 97.0 94.2 0.0123 1 minute k 0.0002 seconds k 625 relative recognition speed, see text for details. y on a Sun Sparc II workstation, p.92 of [9] ....

[Article contains additional citation context not shown here]

S. Lawrence, C. Lee Giles, A. Tsoi, and A. Back, "Face recognition: A convolutional neural network approach," IEEE Transactions on Neural Networks, vol. 8, no. 1, pp. 98--113, 1997.

....[11] repeat the experiments again, and argue that even a simple matching algorithm can deliver nearly the same accuracy as SVMs. Thus, it seems that the advantage of using SVMs is not obvious. It is difficult to discriminate or recognize different persons (hundrends or thousands) by their faces [6] because of the similarity of faces. In this research, we focus on the face recognition problem, and show that the discrimination functions learned by SVMs can give much higher recognition accuracy than the popular standard eigenface approach [15] Eigenfaces are used to represent face images ....

....for classification of the ORL database images. In [14] a hidden Markov model (HMM) based approach is used, and the best model resulted in a 13 error rate. Later, Samaria extends the top down HMM [14] with pseudo two dimensional HMMs [13] and the error rate reduces to 5 . Lawrence et al. [6] takes the convolutional neural network (CNN) approach for the classification of ORL database, and the best error rate reported is 3:83 (in the average of three runs) In our face recognition experiments on the ORL database, we select 200 samples (5 for each individual) randomly as the training ....

[Article contains additional citation context not shown here]

S. Lawrence, C. L. Giles, A. C. Tsoi, and A. D. Back. Face recognition: A convolutional neural network approach. IEEE Trans. Neural Networks, 8:98--113, 1997.

.... Fixed size Convolutional Networks have been applied to many applications, among others: handwriting recognition [LeCun et al. 1990,Martin 1993] as well as machine printed character recognition [Wang and Jean 1993] on line handwriting recognition [Bengio et al. 1995] and face recognition [Lawrence et al. 1997]. Fixed size convolutional networks that share weights along a single temporal dimension are known as Time Delay Neural Networks (TDNNs) and applied widely in speech processing and time series prediction. Variable size convolutional networks, which have applications in object detection and ....

Lawrence, S., Giles, C. L., Tsoi, A. C., and Back, A. D. (1997). Face Recognition: A Convolutional Neural Network Approach. IEEE Transactions on Neural Networks, 8(1):98--113.

....on the ORL database. The probabilistic decision based neural net (PDBNN) results are taken from Lin and Kung [10] Self organising map combined with convolutional neural net (SOM CN) results, together with the results of Eigenface, Top down HMM and Pseudo 2d HMM are taken from Lawrence et al. [11] and Samaria and Harter [12] The humble nearest neighbour classifier actually performs surprisingly well. This is based on a city block distance; a Euclidean distance version performs significantly worse. The question of the statistical significance of these difference arises, and it will be ....

S. Lawrence, C. Giles, A. Tsoi, and A. Back, "Face recognition: a convolutional neural network approach," IEEE Transactions on Neural Networks, vol. 8, pp. 98 -- 113, (1997).

....mode. The auto association NN automatically extracts features (as the output of a hidden layer) that are used by the classification NN. The resulting feature vector is the same as that produced by the eigenface method if the auto association net is linear. More recently, Lawrence et al. [43] proposed a hybrid NN approach that combines local image sampling, a self organizing map (SOP) and a convolutional NN. The SOP provides a a set of features that represents a more compact and robust representation of the image samples. These features are then fed into a convolutional NN. This ....

....subject per subject calculation Auto Association and 40 subjects 40 subjects Not Classification NN [42] 73] 5 images 5images 20 specified per subject per subject PDBNN [45] 40 subjects 40 subjects 0. 1 sec 5 images 5 images 96 on SGI Indy per subject per subject 100 MHz Convolutional NN [43] 40 subjects 40 subjects 0.5 sec 5 images 5 images 96.2 on SGI Indy per subject per subject 100 MHz Dynamic Link 40 subjects 40 subjects Not Matching [51] 73] 5 images 5 images 80 specified per subject per subject VFR [60] 40 subjects 40 subjects 320 sec on 5 images 5 images 92.5 ....

[Article contains additional citation context not shown here]

A. Lawrence, C. Giles, A. Tsoi, and A. Back, "Face recognition : A convolutional neural network approach," IEEE Transactions on Neural Networks, vol. 8, no. 1, pp. 98--113, 1997. 119

....in the image processing and pattern recognition research area. The applications of face recognition systems are manifold, like access control, video surveillance, credit card user identification and automatic video indexing. In recent years many approaches to face recognition have been developed. [3, 6] give an overview of the different face recognition techniques. Popular methods for face recognition are eigenfaces [13] neural networks [6, 7] and bunch graph matching [14] 8] applies a n tuple classifier to face recognition. Several systems use Hidden Markov Models for face recognition [1, 9, ....

....video surveillance, credit card user identification and automatic video indexing. In recent years many approaches to face recognition have been developed. 3, 6] give an overview of the different face recognition techniques. Popular methods for face recognition are eigenfaces [13] neural networks [6, 7] and bunch graph matching [14] 8] applies a n tuple classifier to face recognition. Several systems use Hidden Markov Models for face recognition [1, 9, 11, 12] In this paper we will focus on pseudo 2 D Hidden Markov Model based face recognition systems. Although our system is not based on ....

[Article contains additional citation context not shown here]

S. Lawrence, C. L. Giles, A. C. Tsoi, and A. D. Back. Face Recognition: A Convolutional Neural Network Approach. IEEE Transactions on Neural Networks, 8(1):98--113, 1997.

No context found.

S. Lawrence, C. L. Giles, A. C. Tsoi and A. D. Back, "Face Recognition: A Convolutional Neural Networks Approach," IEEE Trans. on Neural Networks, Special Issue on Neural Networks and Pattern Recognition, Vol. 8, No. 1, pp. 98-113, 1997.

No context found.

S.Lawrence and C.Giles and A.Tsoi and A.Back. "Face Recognition: A Convolutional Neural Network Approach", IEEE Transactions on Neural Networks, vol. 8, no. 1, pp. 98-113, 1998.

No context found.

S. Lawrence et al., "Face Recognition: A Convolutional NeuralNetwork Approach", IEEE Trans. Neural Net., Vol. 8, No. 1, 1997.

No context found.

S. Lawrence, C.L. Giles, A.C. Tsoi, and A.D. Back. Face recognition: A convolutional neural network approach. IEEE Transactions on Neural Networks, 8(1):98-- 113, 1997.

No context found.

Steve Lawrence, C. Lee Giles, A. C. Tsoi, and A. D. Back, (1997)"Face recognition: A convolutional neural network approach", IEEE Transactions on Neural Networks, vol. 8, no. 1, pp. 98-113. Junping Zhang, Stan Z. Li, and Jue Wang

No context found.

S. Lawrence, C. L. Giles, A. Tsoi, and A. Back. "Face recognition: A convolutional neural network approach". IEEE Transactions on Neural Networks, 8(1):98--113, 1997.

No context found.

S. Lawrence, C. L. Giles, A. C. Tsoi, and A. D. Back, "Face recognition: a convolutional neural-network approach," IEEE Trans. Neural Networks, vol. 8, no. 1, pp. 98 --113, 1997. 22

No context found.

Steve Lawrence, C. Lee Giles, A. C. Tsoi, and A. D. Back, (1997)"Face recognition: A convolutional neural network approach", IEEE Transactions on Neural Networks, vol. 8, no. 1, pp. 98-113. Junping Zhang, Stan Z. Li, and Jue Wang

No context found.

S. Lawrence, C. L. Giles, A. Tsoi, and A. Back. "Face recognition: A convolutional neural network approach". IEEE Transactions on Neural Networks, 8(1):98--113, 1997.

No context found.

S Lawrence, C. Giles, A. Tsoi and A. Back, Face recognition: a convolutional neural network approach, IEEE Transactions on Neural Networks, 8 (1) 98-113, 1997.

No context found.

S. Lawrence, C. L. Giles, A. -C. Tsoi, and A. Back, "Face recognition: A convolutional neural network approach," IEEE Trans. Neural Networks, vol. 8, pp. 98--113, Jan. 1997.

No context found.

S.Lawrence,C.L.Giles,A.C.TsoiandA.D.Back, "Face Recognition: A Convolutional Neural Networks Approach", IEEE Trans. on Neural Networks, Special Issue on Neural Networks and Pattern Recognition,Vol. 8, No. 1, pp. 98-113, 1997.

No context found.

S. Lawrence, C.L. Giles, A.C. Tsoi, and A.D. Back. Face recognition: A convolutional neural network approach. IEEE Transactions on Neural Networks, 8(1):98-- 113, 1997.

No context found.

S. Lawrence, C. Giles, A. Tsoi, and A. Back. Face recognition: A convolutional neural network approach. IEEE Trans. Neural Networks, 8:98--113, 1997.

No context found.

S. Lawrence, C. Giles, A. Tsoi, and A. Back. Face recognition: A convolutional neural network approach. IEEE Transactions on Neural Network and Pattern Recognition, 8:98--113, 1997.

No context found.

S. Lawrence, C. L. Giles, A. C. Tsoi and A. D. Back, \ Face Recognition: A Convolutional Neural-Network Approach", IEEE Trans. on Neural Networks, Vol. 8, No. 1, pp. 98-113, 1997.

No context found.

S. Lawrence, C. Lee Giles, A. C. Tsoi, and A. D. Back, "Face recognition: A convolutional neural network approach," IEEE Transactions on Neural Networks, vol. 8, no. 1, pp. 98--113, 1997.

No context found.

S.Lawrence, C. L. Giles, A.C. Tsoi and A. D. Back, "Face Recognition: A Convolutional Neural-Network Approach", IEEE Trans. Neural Networks, vol. 8, no. 1, pp. 98-113, January 1997.

No context found.

Lawrence S; Giles CL; Tsoi AC; Back AD. Face recognition: A convolutional neural-network approach. IEEE Transactions on Neural Networks, 8(1):98 -- 113, 1997. 8 FE Plane Number of classes Correct Rejected LoG 121 1138 (97%) 440 (27%) Gabor 122 1386 (98.6%) 208 (13%)

No context found.

S. Lawrence et al., "Face Recognition: A Convolutional NeuralNetwork Approach", IEEE Trans. Neural Net., Vol. 8, No. 1, 1997.

No context found.

A. Lawrence, C. Giles, A. Tsoi and A. Back, "Face recognition: A convolutional neural network approach," IEEE Transactions on Neural Networks, vol. 8, no. 1, pp. 98-113, 1997.

No context found.

S. Lawrence, C. L. Giles, A. C. Tsoi, and A. D. Back. Face recognition: A convolutional neural network approach. IEEE transactions on Neural Networks, 8(1):98-112, Jan 1997.

No context found.

Steve Lawrence, C. Lee Giles, A.C. Tsoi, and A.D. Back. Face recognition: A convolutional neural network approach. IEEE Transactions on Neural Networks, 8(1):98-113, 1997.

First 50 documents  Next 50

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