2013 IEEE International Conference on Image Processing, 2013
We introduce new approaches for augmenting annotated training datasets used for object detection ... more We introduce new approaches for augmenting annotated training datasets used for object detection tasks that serve achieving two goals: reduce the effort needed for collecting and manually annotating huge datasets and introduce novel variations to the initial dataset that help the learning algorithms. The methods presented in this work aim at relocating objects using their segmentation masks to new backgrounds. These variations comprise changes in properties of objects such as spatial location in the image, surrounding context and scale. We propose a model selection approach to arbitrate between the constructed model on a per class basis. Experimental results show gains that can be harvested using the proposed approach.
2015 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2015
As mobile devices are becoming more ubiquitous, it becomes important to continuously verify the i... more As mobile devices are becoming more ubiquitous, it becomes important to continuously verify the identity of the user during all interactions rather than just at login time. This paper investigates the effectiveness of methods for fully-automatic face recognition in solving the Active Authentication (AA) problem for smartphones. We report the results of face authentication using videos recorded by the front camera. The videos were acquired while the users were performing a number of tasks under three different ambient conditions to capture the type of variations caused by the 'mobility' of the devices. An inspection of these videos reveal a combination of favorable and challenging properties unique to smartphone face videos. In addition to the variations caused by the mobility of the device, other challenges in the dataset include partial faces, occasional pose changes, blur and face/fiducial points localization errors. We evaluate still image and image set-based authentication algorithms using intensity features extracted around fiducial points. The recognition rates drop dramatically when enrollment and test videos come from different sessions. We will make the dataset and the computed features publicly available to help the design of algorithms that are more robust to variations due to factors mentioned above.
ABSTRACT The authors propose a new biometric modality, called a screen fingerprint, for active au... more ABSTRACT The authors propose a new biometric modality, called a screen fingerprint, for active authentication. A screen fingerprint is acquired by taking a screen recording of the computer being used and extracting discriminative visual features from the recording. The authors demonstrate that the screen fingerprint of an operator captures enough unique human qualities to be usable as a biometric for authentication. The also discuss the advantages over other traditional active authentication modalities. This article is part of a special issue on security.
2013 IEEE International Conference on Image Processing, 2013
We introduce new approaches for augmenting annotated training datasets used for object detection ... more We introduce new approaches for augmenting annotated training datasets used for object detection tasks that serve achieving two goals: reduce the effort needed for collecting and manually annotating huge datasets and introduce novel variations to the initial dataset that help the learning algorithms. The methods presented in this work aim at relocating objects using their segmentation masks to new backgrounds. These variations comprise changes in properties of objects such as spatial location in the image, surrounding context and scale. We propose a model selection approach to arbitrate between the constructed model on a per class basis. Experimental results show gains that can be harvested using the proposed approach.
Procedings of the British Machine Vision Conference 2010, 2010
Page 1. FATHY et al.: THE E8P APPROACH FOR FM ESTIMATION 1 Simple, Fast and Accurate Estimation o... more Page 1. FATHY et al.: THE E8P APPROACH FOR FM ESTIMATION 1 Simple, Fast and Accurate Estimation of the Fundamental Matrix Using the Extended Eight-Point Schemes Mohammed E. Fathy mefathy(at)gmail.com Ashraf ...
Screen touch gesture has been shown to be a promising modality for touch-based active authenticat... more Screen touch gesture has been shown to be a promising modality for touch-based active authentication of users of mobile devices. In this paper, we present an approach for active user authentication using screen touch gestures by building linear and kernelized dictionaries based on sparse representations and associated classifiers. Experiments using a new dataset collected by us as well as two other publicly available screen touch datasets show that the dictionary-based classification method compares favorably to those published in the literature. Experiments done using data collected in three different sessions corresponding to different environmental conditions show a drop in performance when the training and test data come from different sessions. This suggests a need for applying domain adaptation methods to further improve the performance of the classifiers.
As mobile devices are becoming more ubiquitous, it becomes important to continuously verify the i... more As mobile devices are becoming more ubiquitous, it becomes important to continuously verify the identity of the user during all interactions rather than just at login time. This paper investigates the effectiveness of methods for fully-automatic face recognition in solving the Active Authentication (AA) problem for smartphones. We report the results of face authentication using videos recorded by the front camera. The videos were acquired while the users were performing a number of tasks under three different ambient conditions to capture the type of variations caused by the 'mobility' of the devices. An inspection of these videos reveal a combination of favorable and challenging properties unique to smartphone face videos. In addition to the variations caused by the mobility of the device, other challenges in the dataset include partial faces, occasional pose changes, blur and face/fiducial points localization errors. We evaluate still image and image set-based authentication algorithms using intensity features extracted around fiducial points. The recognition rates drop dramatically when enrollment and test videos come from different sessions. We will make the dataset and the computed features publicly available to help the design of algorithms that are more robust to variations due to factors mentioned above.
We investigate if screen-based recordings of computer interactions can be used for accurate activ... more We investigate if screen-based recordings of computer interactions can be used for accurate active user authentication. A dataset of screen recordings of some PC interactions (MouseMoving, Typing, Scrolling, Other) of 21 users was collected and we ran a set of experiments to help our investigation. Low-dimensional feature vectors based on histogram of optical flows from each screen recording were used in our study. The first set of experiments investigated if these low-dimensional features can be used to recognize the type of interaction taking place in a particular recording and we found that linear SVM could succeed in achieving this with an accuracy of 91% on 5 test users. The second set of experiments explored if classifiers trained on different types of recordings can be used to verify user identity. The results indicated that SVMs trained on Scrolling recordings can achieve moderately low FAR and FRR error rates of 20.7% and 12.4%, respectively. These preliminary results indicate that further research in using screen-based recordings for active authentication can lead to a reliable soft cyber biometric.
The fundamental matrix (FM) describes the geometric relations that exist be-1 tween two images of... more The fundamental matrix (FM) describes the geometric relations that exist be-1 tween two images of the same scene. Different error criteria are used for esti-2 mating FMs from an input set of correspondences. In this paper, the accuracy 3 and efficiency aspects of the different error criteria were studied. We mathe-4 matically and experimentally proved that the most popular error criterion, the 5 symmetric epipolar distance, is biased. It was also shown that despite the simi-6 larity between the algebraic expressions of the symmetric epipolar distance and 7 Sampson distance, they have different accuracy properties. In addition, a new 8 error criterion, Kanatani distance, was proposed and was proved to be the most 9 effective for use during the outlier removal phase from accuracy and efficiency 10 perspectives. To thoroughly test the accuracy of the different error criteria, we 11 proposed a randomized algorithm for Reprojection Error-based Correspondence 12 Generation (RE-CG). As input, RE-CG takes an FM and a desired reprojection 13 error value d. As output, RE-CG generates a random correspondence having 14 that error value. Mathematical analysis of this algorithm revealed that the suc-15 cess probability for any given trial is 1 − (2/3) 2 at best and is 1 − (6/7) 2 at 16 worst while experiments demonstrated that the algorithm often succeeds after 17 only one trial. 18 The fundamental matrix (FM) relating two images (I, I ) is estimated from 21 a number of correspondences between I and I . A correspondence is a pair 22 of points (p, p ) on the two images (I, I ) that are believed to be projections 23 of the same 3D point. Automatic algorithms for identifying correspondences 24 not only introduce errors in the computed locations of the points (localization 25 errors), but also produce totally false matches (outliers) (Zhang and Kanade, 26 1998). To get acceptable results, FM estimation usually starts by removing 27 these outliers. Then, a one-step FM estimation technique such as the eight-28 point algorithm (Hartley and Zisserman, 2004; Zhang and Kanade, 1998) is 29 used to obtain a better estimate of the FM by taking into account the effect 30 of all the inliers rather than just a small, 7-point subset. Finally, the result 31 obtained by the one-step method is refined using an iterative technique. 32 FM error criteria play a vital role in the process of the FM estimation. 33 An FM error criterion is a real-valued function that measures the amount of 34 deviation of a given correspondence from the epipolar constraint parameterized 35 by a given FM. FM error criteria are used in three different situations: 36 1. During the outlier removal phase of the FM estimation, an error criterion 37 is used as a distance function to measure the proximity of each correspon-38 dence to the current FM hypothesis. 39 2. During the iterative refinement of the FM, an error criterion is used as a 40 cost function to be minimized over the space of rank-2 matrices.
2013 IEEE International Conference on Image Processing, 2013
We introduce new approaches for augmenting annotated training datasets used for object detection ... more We introduce new approaches for augmenting annotated training datasets used for object detection tasks that serve achieving two goals: reduce the effort needed for collecting and manually annotating huge datasets and introduce novel variations to the initial dataset that help the learning algorithms. The methods presented in this work aim at relocating objects using their segmentation masks to new backgrounds. These variations comprise changes in properties of objects such as spatial location in the image, surrounding context and scale. We propose a model selection approach to arbitrate between the constructed model on a per class basis. Experimental results show gains that can be harvested using the proposed approach.
2015 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2015
As mobile devices are becoming more ubiquitous, it becomes important to continuously verify the i... more As mobile devices are becoming more ubiquitous, it becomes important to continuously verify the identity of the user during all interactions rather than just at login time. This paper investigates the effectiveness of methods for fully-automatic face recognition in solving the Active Authentication (AA) problem for smartphones. We report the results of face authentication using videos recorded by the front camera. The videos were acquired while the users were performing a number of tasks under three different ambient conditions to capture the type of variations caused by the 'mobility' of the devices. An inspection of these videos reveal a combination of favorable and challenging properties unique to smartphone face videos. In addition to the variations caused by the mobility of the device, other challenges in the dataset include partial faces, occasional pose changes, blur and face/fiducial points localization errors. We evaluate still image and image set-based authentication algorithms using intensity features extracted around fiducial points. The recognition rates drop dramatically when enrollment and test videos come from different sessions. We will make the dataset and the computed features publicly available to help the design of algorithms that are more robust to variations due to factors mentioned above.
ABSTRACT The authors propose a new biometric modality, called a screen fingerprint, for active au... more ABSTRACT The authors propose a new biometric modality, called a screen fingerprint, for active authentication. A screen fingerprint is acquired by taking a screen recording of the computer being used and extracting discriminative visual features from the recording. The authors demonstrate that the screen fingerprint of an operator captures enough unique human qualities to be usable as a biometric for authentication. The also discuss the advantages over other traditional active authentication modalities. This article is part of a special issue on security.
2013 IEEE International Conference on Image Processing, 2013
We introduce new approaches for augmenting annotated training datasets used for object detection ... more We introduce new approaches for augmenting annotated training datasets used for object detection tasks that serve achieving two goals: reduce the effort needed for collecting and manually annotating huge datasets and introduce novel variations to the initial dataset that help the learning algorithms. The methods presented in this work aim at relocating objects using their segmentation masks to new backgrounds. These variations comprise changes in properties of objects such as spatial location in the image, surrounding context and scale. We propose a model selection approach to arbitrate between the constructed model on a per class basis. Experimental results show gains that can be harvested using the proposed approach.
Procedings of the British Machine Vision Conference 2010, 2010
Page 1. FATHY et al.: THE E8P APPROACH FOR FM ESTIMATION 1 Simple, Fast and Accurate Estimation o... more Page 1. FATHY et al.: THE E8P APPROACH FOR FM ESTIMATION 1 Simple, Fast and Accurate Estimation of the Fundamental Matrix Using the Extended Eight-Point Schemes Mohammed E. Fathy mefathy(at)gmail.com Ashraf ...
Screen touch gesture has been shown to be a promising modality for touch-based active authenticat... more Screen touch gesture has been shown to be a promising modality for touch-based active authentication of users of mobile devices. In this paper, we present an approach for active user authentication using screen touch gestures by building linear and kernelized dictionaries based on sparse representations and associated classifiers. Experiments using a new dataset collected by us as well as two other publicly available screen touch datasets show that the dictionary-based classification method compares favorably to those published in the literature. Experiments done using data collected in three different sessions corresponding to different environmental conditions show a drop in performance when the training and test data come from different sessions. This suggests a need for applying domain adaptation methods to further improve the performance of the classifiers.
As mobile devices are becoming more ubiquitous, it becomes important to continuously verify the i... more As mobile devices are becoming more ubiquitous, it becomes important to continuously verify the identity of the user during all interactions rather than just at login time. This paper investigates the effectiveness of methods for fully-automatic face recognition in solving the Active Authentication (AA) problem for smartphones. We report the results of face authentication using videos recorded by the front camera. The videos were acquired while the users were performing a number of tasks under three different ambient conditions to capture the type of variations caused by the 'mobility' of the devices. An inspection of these videos reveal a combination of favorable and challenging properties unique to smartphone face videos. In addition to the variations caused by the mobility of the device, other challenges in the dataset include partial faces, occasional pose changes, blur and face/fiducial points localization errors. We evaluate still image and image set-based authentication algorithms using intensity features extracted around fiducial points. The recognition rates drop dramatically when enrollment and test videos come from different sessions. We will make the dataset and the computed features publicly available to help the design of algorithms that are more robust to variations due to factors mentioned above.
We investigate if screen-based recordings of computer interactions can be used for accurate activ... more We investigate if screen-based recordings of computer interactions can be used for accurate active user authentication. A dataset of screen recordings of some PC interactions (MouseMoving, Typing, Scrolling, Other) of 21 users was collected and we ran a set of experiments to help our investigation. Low-dimensional feature vectors based on histogram of optical flows from each screen recording were used in our study. The first set of experiments investigated if these low-dimensional features can be used to recognize the type of interaction taking place in a particular recording and we found that linear SVM could succeed in achieving this with an accuracy of 91% on 5 test users. The second set of experiments explored if classifiers trained on different types of recordings can be used to verify user identity. The results indicated that SVMs trained on Scrolling recordings can achieve moderately low FAR and FRR error rates of 20.7% and 12.4%, respectively. These preliminary results indicate that further research in using screen-based recordings for active authentication can lead to a reliable soft cyber biometric.
The fundamental matrix (FM) describes the geometric relations that exist be-1 tween two images of... more The fundamental matrix (FM) describes the geometric relations that exist be-1 tween two images of the same scene. Different error criteria are used for esti-2 mating FMs from an input set of correspondences. In this paper, the accuracy 3 and efficiency aspects of the different error criteria were studied. We mathe-4 matically and experimentally proved that the most popular error criterion, the 5 symmetric epipolar distance, is biased. It was also shown that despite the simi-6 larity between the algebraic expressions of the symmetric epipolar distance and 7 Sampson distance, they have different accuracy properties. In addition, a new 8 error criterion, Kanatani distance, was proposed and was proved to be the most 9 effective for use during the outlier removal phase from accuracy and efficiency 10 perspectives. To thoroughly test the accuracy of the different error criteria, we 11 proposed a randomized algorithm for Reprojection Error-based Correspondence 12 Generation (RE-CG). As input, RE-CG takes an FM and a desired reprojection 13 error value d. As output, RE-CG generates a random correspondence having 14 that error value. Mathematical analysis of this algorithm revealed that the suc-15 cess probability for any given trial is 1 − (2/3) 2 at best and is 1 − (6/7) 2 at 16 worst while experiments demonstrated that the algorithm often succeeds after 17 only one trial. 18 The fundamental matrix (FM) relating two images (I, I ) is estimated from 21 a number of correspondences between I and I . A correspondence is a pair 22 of points (p, p ) on the two images (I, I ) that are believed to be projections 23 of the same 3D point. Automatic algorithms for identifying correspondences 24 not only introduce errors in the computed locations of the points (localization 25 errors), but also produce totally false matches (outliers) (Zhang and Kanade, 26 1998). To get acceptable results, FM estimation usually starts by removing 27 these outliers. Then, a one-step FM estimation technique such as the eight-28 point algorithm (Hartley and Zisserman, 2004; Zhang and Kanade, 1998) is 29 used to obtain a better estimate of the FM by taking into account the effect 30 of all the inliers rather than just a small, 7-point subset. Finally, the result 31 obtained by the one-step method is refined using an iterative technique. 32 FM error criteria play a vital role in the process of the FM estimation. 33 An FM error criterion is a real-valued function that measures the amount of 34 deviation of a given correspondence from the epipolar constraint parameterized 35 by a given FM. FM error criteria are used in three different situations: 36 1. During the outlier removal phase of the FM estimation, an error criterion 37 is used as a distance function to measure the proximity of each correspon-38 dence to the current FM hypothesis. 39 2. During the iterative refinement of the FM, an error criterion is used as a 40 cost function to be minimized over the space of rank-2 matrices.
Uploads
Papers by Mohammed Fathy