Pub Date : 2015-05-19DOI: 10.1109/ICB.2015.7139070
Jianqing Zhu, Shengcai Liao, Dong Yi, Zhen Lei, S. Li
Recently, pedestrian attributes like gender, age and clothing etc., have been used as soft biometric traits for recognizing people. Unlike existing methods that assume the independence of attributes during their prediction, we propose a multi-label convolutional neural network (MLCNN) to predict multiple attributes together in a unified framework. Firstly, a pedestrian image is roughly divided into multiple overlapping body parts, which are simultaneously integrated in the multi-label convolutional neural network. Secondly, these parts are filtered independently and aggregated in the cost layer. The cost function is a combination of multiple binary attribute classification cost functions. Moreover, we propose an attribute assisted person re-identification method, which fuses attribute distances and low-level feature distances between pairs of person images to improve person re-identification performance. Extensive experiments show: 1) the average attribute classification accuracy of the proposed method is 5.2% and 9.3% higher than the SVM-based method on three public databases, VIPeR and GRID, respectively; 2) the proposed attribute assisted person re-identification method is superior to existing approaches.
{"title":"Multi-label CNN based pedestrian attribute learning for soft biometrics","authors":"Jianqing Zhu, Shengcai Liao, Dong Yi, Zhen Lei, S. Li","doi":"10.1109/ICB.2015.7139070","DOIUrl":"https://doi.org/10.1109/ICB.2015.7139070","url":null,"abstract":"Recently, pedestrian attributes like gender, age and clothing etc., have been used as soft biometric traits for recognizing people. Unlike existing methods that assume the independence of attributes during their prediction, we propose a multi-label convolutional neural network (MLCNN) to predict multiple attributes together in a unified framework. Firstly, a pedestrian image is roughly divided into multiple overlapping body parts, which are simultaneously integrated in the multi-label convolutional neural network. Secondly, these parts are filtered independently and aggregated in the cost layer. The cost function is a combination of multiple binary attribute classification cost functions. Moreover, we propose an attribute assisted person re-identification method, which fuses attribute distances and low-level feature distances between pairs of person images to improve person re-identification performance. Extensive experiments show: 1) the average attribute classification accuracy of the proposed method is 5.2% and 9.3% higher than the SVM-based method on three public databases, VIPeR and GRID, respectively; 2) the proposed attribute assisted person re-identification method is superior to existing approaches.","PeriodicalId":237372,"journal":{"name":"2015 International Conference on Biometrics (ICB)","volume":"2 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2015-05-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"120953405","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2015-05-19DOI: 10.1109/ICB.2015.7139082
Keyurkumar Patel, Hu Han, Anil K. Jain, Greg Ott
With the wide deployment of face recognition systems in applications from border control to mobile device unlocking, the combat of face spoofing attacks requires increased attention; such attacks can be easily launched via printed photos, video replays and 3D masks. We address the problem of facial spoofing detection against replay attacks based on the analysis of aliasing in spoof face videos. The application domain of interest is mobile phone unlock. We analyze the moiré pattern aliasing that commonly appears during the recapture of video or photo replays on a screen in different channels (R, G, B and grayscale) and regions (the whole frame, detected face, and facial component between the nose and chin). Multi-scale LBP and DSIFT features are used to represent the characteristics of moiré patterns that differentiate a replayed spoof face from a live face (face present). Experimental results on Idiap replay-attack and CASIA databases as well as a database collected in our laboratory (RAFS), which is based on the MSU-FSD database, shows that the proposed approach is very effective in face spoof detection for both cross-database, and intra-database testing scenarios.
{"title":"Live face video vs. spoof face video: Use of moiré patterns to detect replay video attacks","authors":"Keyurkumar Patel, Hu Han, Anil K. Jain, Greg Ott","doi":"10.1109/ICB.2015.7139082","DOIUrl":"https://doi.org/10.1109/ICB.2015.7139082","url":null,"abstract":"With the wide deployment of face recognition systems in applications from border control to mobile device unlocking, the combat of face spoofing attacks requires increased attention; such attacks can be easily launched via printed photos, video replays and 3D masks. We address the problem of facial spoofing detection against replay attacks based on the analysis of aliasing in spoof face videos. The application domain of interest is mobile phone unlock. We analyze the moiré pattern aliasing that commonly appears during the recapture of video or photo replays on a screen in different channels (R, G, B and grayscale) and regions (the whole frame, detected face, and facial component between the nose and chin). Multi-scale LBP and DSIFT features are used to represent the characteristics of moiré patterns that differentiate a replayed spoof face from a live face (face present). Experimental results on Idiap replay-attack and CASIA databases as well as a database collected in our laboratory (RAFS), which is based on the MSU-FSD database, shows that the proposed approach is very effective in face spoof detection for both cross-database, and intra-database testing scenarios.","PeriodicalId":237372,"journal":{"name":"2015 International Conference on Biometrics (ICB)","volume":"24 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2015-05-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121346906","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2015-05-19DOI: 10.1109/ICB.2015.7139084
Christof Kauba, A. Uhl
The impact of sensor ageing related pixel defects on the performance of finger vein based recognition systems in terms of the EER (Equal Error Rate) is investigated. Therefore the defect growth rate per year for the sensor used to capture the data set was estimated. Based on this estimation an experimental study using several simulations with increasing numbers of stuck and hot pixels were done to determine the impact on different finger-vein matching schemes. Whereas for a reasonable number of pixel defects none of the methods is considerably influenced, the performance of several schemes drops if the number of defects is increased. The impact can be reduced using a simple denoising filter.
{"title":"Sensor ageing impact on finger-vein recognition","authors":"Christof Kauba, A. Uhl","doi":"10.1109/ICB.2015.7139084","DOIUrl":"https://doi.org/10.1109/ICB.2015.7139084","url":null,"abstract":"The impact of sensor ageing related pixel defects on the performance of finger vein based recognition systems in terms of the EER (Equal Error Rate) is investigated. Therefore the defect growth rate per year for the sensor used to capture the data set was estimated. Based on this estimation an experimental study using several simulations with increasing numbers of stuck and hot pixels were done to determine the impact on different finger-vein matching schemes. Whereas for a reasonable number of pixel defects none of the methods is considerably influenced, the performance of several schemes drops if the number of defects is increased. The impact can be reduced using a simple denoising filter.","PeriodicalId":237372,"journal":{"name":"2015 International Conference on Biometrics (ICB)","volume":"20 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2015-05-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121393443","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2015-05-19DOI: 10.1109/ICB.2015.7139048
D. Muramatsu, Yasushi Makihara, Y. Yagi
Gait recognition has potential to recognize subject in CCTV footages thanks to robustness against image resolution. In the CCTV footage, several body-regions of subjects are, however, often un-observable because of occlusions and/or cutting off caused by limited field of view, and therefore, recognition must be done from a pair of partially observed data. The most popular approach to recognition from partially observed data is matching the data from common observable region. This approach, however, cannot be applied in the cases where the matching pair has no common observable region. We therefore, propose an approach to enable recognition even from the pair with no common observable region. In the proposed approach, we reconstruct entire gait feature from a partial gait feature extracted from the observable region using a subspace-based method, and match the reconstructed entire gait features for recognition. We evaluate the proposed approach against two different datasets. In the best case, the proposed approach achieves recognition accuracy with EER of 16.2% from such a pair.
{"title":"Gait regeneration for recognition","authors":"D. Muramatsu, Yasushi Makihara, Y. Yagi","doi":"10.1109/ICB.2015.7139048","DOIUrl":"https://doi.org/10.1109/ICB.2015.7139048","url":null,"abstract":"Gait recognition has potential to recognize subject in CCTV footages thanks to robustness against image resolution. In the CCTV footage, several body-regions of subjects are, however, often un-observable because of occlusions and/or cutting off caused by limited field of view, and therefore, recognition must be done from a pair of partially observed data. The most popular approach to recognition from partially observed data is matching the data from common observable region. This approach, however, cannot be applied in the cases where the matching pair has no common observable region. We therefore, propose an approach to enable recognition even from the pair with no common observable region. In the proposed approach, we reconstruct entire gait feature from a partial gait feature extracted from the observable region using a subspace-based method, and match the reconstructed entire gait features for recognition. We evaluate the proposed approach against two different datasets. In the best case, the proposed approach achieves recognition accuracy with EER of 16.2% from such a pair.","PeriodicalId":237372,"journal":{"name":"2015 International Conference on Biometrics (ICB)","volume":"2 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2015-05-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121089900","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2015-05-19DOI: 10.1109/ICB.2015.7139079
Junlin Hu, Jiwen Lu, Yap-Peng Tan
This paper investigates the problem of fine-grained face verification under unconstrained conditions. For the conventional face verification task, the verification model is trained with some positive and negative face pairs, where each positive sample pair contains two face images of the same person while each negative sample pair usually consists of two face images from different subjects. However, in many real applications, facial appearance of the twins looks very similar even if they are considered as a negative pair in face verification. Therefore, it is important to differentiate a given face pair to determine whether it is from the same person or a twins for a practical face verification system because most existing face verification systems fails to work well in such a scenario. In this work, we define the problem as fine-grained face verification and collect an unconstrained face dataset which contains 455 pairs of identical twins to generate negative face pairs to evaluate several baseline verification models for fine-grained unconstrained face verification. Benchmark results on the unsupervised setting and restricted setting show the challenge of the fine-grained face verification in the wild.
{"title":"Fine-grained face verification: Dataset and baseline results","authors":"Junlin Hu, Jiwen Lu, Yap-Peng Tan","doi":"10.1109/ICB.2015.7139079","DOIUrl":"https://doi.org/10.1109/ICB.2015.7139079","url":null,"abstract":"This paper investigates the problem of fine-grained face verification under unconstrained conditions. For the conventional face verification task, the verification model is trained with some positive and negative face pairs, where each positive sample pair contains two face images of the same person while each negative sample pair usually consists of two face images from different subjects. However, in many real applications, facial appearance of the twins looks very similar even if they are considered as a negative pair in face verification. Therefore, it is important to differentiate a given face pair to determine whether it is from the same person or a twins for a practical face verification system because most existing face verification systems fails to work well in such a scenario. In this work, we define the problem as fine-grained face verification and collect an unconstrained face dataset which contains 455 pairs of identical twins to generate negative face pairs to evaluate several baseline verification models for fine-grained unconstrained face verification. Benchmark results on the unsupervised setting and restricted setting show the challenge of the fine-grained face verification in the wild.","PeriodicalId":237372,"journal":{"name":"2015 International Conference on Biometrics (ICB)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2015-05-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122547733","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2015-05-19DOI: 10.1109/ICB.2015.7139051
Jing Li, Shuqin Long, Dan Zeng, Qijun Zhao
Reconstructing 3D face models from multiple uncalibrated 2D face images is usually done by using a single reference 3D face model or some gender/ethnicity-specific 3D face models. However, different persons, even those of the same gender or ethnicity, usually have significantly different faces in terms of their overall appearance, which forms the base of person recognition using faces. Consequently, existing 3D reference model based methods have limited capability of reconstructing 3D face models for a large variety of persons. In this paper, we propose to explore a reservoir of diverse reference models to improve the 3D face reconstruction performance. Specifically, we convert the face reconstruction problem into a multi-label segmentation problem. Its energy function is formulated from different cues, including 1) similarity between the desired output and the initial model, 2) color consistency between different views, 3) smoothness constraint on adjacent pixels, and 4) model consistency within local neighborhood. Experimental results on challenging datasets demonstrate that the proposed algorithm is capable of recovering high quality face models in both qualitative and quantitative evaluations.
{"title":"Example-based 3D face reconstruction from uncalibrated frontal and profile images","authors":"Jing Li, Shuqin Long, Dan Zeng, Qijun Zhao","doi":"10.1109/ICB.2015.7139051","DOIUrl":"https://doi.org/10.1109/ICB.2015.7139051","url":null,"abstract":"Reconstructing 3D face models from multiple uncalibrated 2D face images is usually done by using a single reference 3D face model or some gender/ethnicity-specific 3D face models. However, different persons, even those of the same gender or ethnicity, usually have significantly different faces in terms of their overall appearance, which forms the base of person recognition using faces. Consequently, existing 3D reference model based methods have limited capability of reconstructing 3D face models for a large variety of persons. In this paper, we propose to explore a reservoir of diverse reference models to improve the 3D face reconstruction performance. Specifically, we convert the face reconstruction problem into a multi-label segmentation problem. Its energy function is formulated from different cues, including 1) similarity between the desired output and the initial model, 2) color consistency between different views, 3) smoothness constraint on adjacent pixels, and 4) model consistency within local neighborhood. Experimental results on challenging datasets demonstrate that the proposed algorithm is capable of recovering high quality face models in both qualitative and quantitative evaluations.","PeriodicalId":237372,"journal":{"name":"2015 International Conference on Biometrics (ICB)","volume":"40 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2015-05-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128562594","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2015-05-19DOI: 10.1109/ICB.2015.7139112
Dayong Wang, Anil K. Jain
Face retrieval is an enabling technology for many applications, including automatic face annotation, deduplication, and surveillance. In this paper, we propose a face retrieval system which combines a k-NN search procedure with a COTS matcher (PittPatt1) in a cascaded manner. In particular, given a query face, we first pre-filter the gallery set and find the top-k most similar faces for the query image by using deep facial features that are learned with a deep convolutional neural network. The top-k most similar faces are then re-ranked based on score-level fusion of the similarities between deep features and the COTS matcher. To further boost the retrieval performance, we develop a manifold ranking algorithm. The proposed face retrieval system is evaluated on two large-scale face image databases: (i) a web face image database, which consists of over 3, 880 query images of 1, 507 subjects and a gallery of 5, 000, 000 faces, and (ii) a mugshot database, which consists of 1, 000 query images of 1, 000 subjects and a gallery of 1, 000, 000 faces. Experimental results demonstrate that the proposed face retrieval system can simultaneously improve the retrieval performance (CMC and precision-recall) and scalability for large-scale face retrieval problems.
{"title":"Face retriever: Pre-filtering the gallery via deep neural net","authors":"Dayong Wang, Anil K. Jain","doi":"10.1109/ICB.2015.7139112","DOIUrl":"https://doi.org/10.1109/ICB.2015.7139112","url":null,"abstract":"Face retrieval is an enabling technology for many applications, including automatic face annotation, deduplication, and surveillance. In this paper, we propose a face retrieval system which combines a k-NN search procedure with a COTS matcher (PittPatt1) in a cascaded manner. In particular, given a query face, we first pre-filter the gallery set and find the top-k most similar faces for the query image by using deep facial features that are learned with a deep convolutional neural network. The top-k most similar faces are then re-ranked based on score-level fusion of the similarities between deep features and the COTS matcher. To further boost the retrieval performance, we develop a manifold ranking algorithm. The proposed face retrieval system is evaluated on two large-scale face image databases: (i) a web face image database, which consists of over 3, 880 query images of 1, 507 subjects and a gallery of 5, 000, 000 faces, and (ii) a mugshot database, which consists of 1, 000 query images of 1, 000 subjects and a gallery of 1, 000, 000 faces. Experimental results demonstrate that the proposed face retrieval system can simultaneously improve the retrieval performance (CMC and precision-recall) and scalability for large-scale face retrieval problems.","PeriodicalId":237372,"journal":{"name":"2015 International Conference on Biometrics (ICB)","volume":"60 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2015-05-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128144388","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2015-05-19DOI: 10.1109/ICB.2015.7139043
David Crouse, Hu Han, Deepak Chandra, Brandon Barbello, Anil K. Jain
Mobile devices can carry large amounts of personal data, but are often left unsecured. PIN locks are inconvenient to use and thus have seen low adoption (33% of users). While biometrics are beginning to be used for mobile device authentication, they are used only for initial unlock. Mobile devices secured with only login authentication are still vulnerable to data theft when in an unlocked state. This paper introduces our work on a face-based continuous authentication system that operates in an unobtrusive manner. We present a methodology for fusing mobile device (unconstrained) face capture with gyroscope, accelerometer, and magnetometer data to correct for camera orientation and, by extension, the orientation of the face image. Experiments demonstrate (i) improvement of face recognition accuracy from face orientation correction, and (ii) efficacy of the prototype continuous authentication system.
{"title":"Continuous authentication of mobile user: Fusion of face image and inertial Measurement Unit data","authors":"David Crouse, Hu Han, Deepak Chandra, Brandon Barbello, Anil K. Jain","doi":"10.1109/ICB.2015.7139043","DOIUrl":"https://doi.org/10.1109/ICB.2015.7139043","url":null,"abstract":"Mobile devices can carry large amounts of personal data, but are often left unsecured. PIN locks are inconvenient to use and thus have seen low adoption (33% of users). While biometrics are beginning to be used for mobile device authentication, they are used only for initial unlock. Mobile devices secured with only login authentication are still vulnerable to data theft when in an unlocked state. This paper introduces our work on a face-based continuous authentication system that operates in an unobtrusive manner. We present a methodology for fusing mobile device (unconstrained) face capture with gyroscope, accelerometer, and magnetometer data to correct for camera orientation and, by extension, the orientation of the face image. Experiments demonstrate (i) improvement of face recognition accuracy from face orientation correction, and (ii) efficacy of the prototype continuous authentication system.","PeriodicalId":237372,"journal":{"name":"2015 International Conference on Biometrics (ICB)","volume":"31 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2015-05-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128928457","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2015-05-19DOI: 10.1109/ICB.2015.7139045
Kamlesh Tiwari, Phalguni Gupta
Fingerphoto is an image of a human finger obtained with the help of an ordinary camera. Its acquisition is convenient and does not require any particular biometric scanner. The high degree of freedom in finger positioning introduces challenges to its recognition. This paper proposes a fingerphoto based human authentication system for mobile hand-held devices by using a non-conventional scale-invariant features. The system utilizes built-in camera of the mobile devices to acquire biometric sample and therefore, eliminates the dependence over specifics scanners. It can successfully handle some of the issues like orientation, rotation and lack of registration at the time of matching. It has achieved CRR of 96.67% and EER of 3.33% which is better than any other system available in the literature.
{"title":"A touch-less fingerphoto recognition system for mobile hand-held devices","authors":"Kamlesh Tiwari, Phalguni Gupta","doi":"10.1109/ICB.2015.7139045","DOIUrl":"https://doi.org/10.1109/ICB.2015.7139045","url":null,"abstract":"Fingerphoto is an image of a human finger obtained with the help of an ordinary camera. Its acquisition is convenient and does not require any particular biometric scanner. The high degree of freedom in finger positioning introduces challenges to its recognition. This paper proposes a fingerphoto based human authentication system for mobile hand-held devices by using a non-conventional scale-invariant features. The system utilizes built-in camera of the mobile devices to acquire biometric sample and therefore, eliminates the dependence over specifics scanners. It can successfully handle some of the issues like orientation, rotation and lack of registration at the time of matching. It has achieved CRR of 96.67% and EER of 3.33% which is better than any other system available in the literature.","PeriodicalId":237372,"journal":{"name":"2015 International Conference on Biometrics (ICB)","volume":"32 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2015-05-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128929685","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2015-05-19DOI: 10.1109/ICB.2015.7139097
Yi Jin, Jiwen Lu, Q. Ruan
This paper presents a Large Margin Coupled Feature Learning (LMCFL) method for cross-modal face recognition, which recognizes persons from facial images captured from different modalities. Most previous cross-modal face recognition methods utilize hand-crafted feature descriptors for face representation, which require strong prior knowledge to engineer and cannot exploit data-adaptive characteristics in feature extraction. In this work, we propose a new LMCFL method to learn coupled face representation at the image pixel level by jointly utilizing the discriminative information of face images in each modality and the correlation information of face images from different modalities. Thus, LMCFL can maximize the margin between positive face pairs and negative face pairs in each modality, and maximize the correlation of face images from different modalities, where discriminative face features can be automatically learned in a discriminative and data-driven way. Our LMCFL is validated on two different cross-modal face recognition applications, and the experimental results demonstrate the effectiveness of our proposed approach.
{"title":"Large Margin Coupled Feature Learning for cross-modal face recognition","authors":"Yi Jin, Jiwen Lu, Q. Ruan","doi":"10.1109/ICB.2015.7139097","DOIUrl":"https://doi.org/10.1109/ICB.2015.7139097","url":null,"abstract":"This paper presents a Large Margin Coupled Feature Learning (LMCFL) method for cross-modal face recognition, which recognizes persons from facial images captured from different modalities. Most previous cross-modal face recognition methods utilize hand-crafted feature descriptors for face representation, which require strong prior knowledge to engineer and cannot exploit data-adaptive characteristics in feature extraction. In this work, we propose a new LMCFL method to learn coupled face representation at the image pixel level by jointly utilizing the discriminative information of face images in each modality and the correlation information of face images from different modalities. Thus, LMCFL can maximize the margin between positive face pairs and negative face pairs in each modality, and maximize the correlation of face images from different modalities, where discriminative face features can be automatically learned in a discriminative and data-driven way. Our LMCFL is validated on two different cross-modal face recognition applications, and the experimental results demonstrate the effectiveness of our proposed approach.","PeriodicalId":237372,"journal":{"name":"2015 International Conference on Biometrics (ICB)","volume":"41 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2015-05-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"117349808","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：