Идентификация лиц в реальном времени с использованием свёрточной нейронной сети и хэширующего леса
Автор: Визильтер Юрий Валентинович, Горбацевич Владимир Сергеевич, Воротников Андрей Валерьевич, Костромов Никита Алексеевич
Журнал: Компьютерная оптика @computer-optics
Рубрика: Обработка изображений: Распознавание образов
Статья в выпуске: 2 т.41, 2017 года.
Бесплатный доступ
В работе предлагается новый подход к построению биометрического шаблона с использованием свёрточной нейронной сети и хэширующего леса. Метод состоит из двух этапов: на первом происходит обучение свёрточной нейронной сети, далее к полученным описаниям применяется хеширующее преобразование с использованием нового предложенного метода хэширующего леса. Такой способ формирования шаблона является обобщением метода Boosted SSC ( Similarity Sensitive Coding ) для решения задачи построения оптимального хеша, учитывающего специфику задач верификации и идентификации лиц. Обучение производилось на базе лиц CASIA-WebFace, а тестирование - на базе лиц LFW. Применительно к задачам бинарного хеширования с метрикой Хемминга описанный подход позволяет получать 200-битный (25 байт) биометрический шаблон с качеством верификации 96,3 % и 2000-битный шаблон с 98,14 % на базе изображений лиц LFW. При использовании хэширующего леса с 7-битными деревьями 2000´7 достигается уровень идентификации в 93 % относительно базовых показателей свёрточной нейронной сети в 89,9%. В результате достигается скорость формирования биометрических шаблонов (описаний лиц) с частотой более 40 изображений лиц в секунду на CPU Core i7 и более 120 изображений лиц в секунду с использованием GPU GeForce GTX 650.
Свёрточные нейронные сети, хэширование, бинарные деревья, метрика хэмминга, биометрия
Короткий адрес: https://sciup.org/14059558
IDR: 14059558 | DOI: 10.18287/2412-6179-2017-41-2-254-265
Real-time face identification via CNN and boosted hashing forest
This paper presents a new approach to constructing a biometric template using a Convolutional Neural Network (CNN) with Hashing Forest. The approach consists of several steps: training a convolutional neural network, transforming it to a multiple convolution architecture, and finally learning the output hashing transform via a new Boosted Hashing Forest technique. This technique generalizes the Boosted SSC (Similarity Sensitive Coding) approach for hashing learning with joint optimization of face verification and identification. The proposed network via hashing forest is trained on the CASIA-WebFace dataset and evaluated on the LFW dataset. The result of coding the output of a single CNN is 97% on LFW. For Hamming embedding, the proposed approach enables a 200 bit (25 byte) code to be constructed with a 96.3% verification accuracy and a 2000-bit code with a 98.14% verification accuracy on LFW. The convolutional network with hashing forest with 2000´7-bit hashing trees achieves 93% rank-1 on LFW relative to the basic convolutional network's 89.9% rank-1. The proposed approach generates templates at the rate of 40+ fps with a GPU Core i7 and 120+ fps with a GPU GeForce GTX 650.
Список литературы Идентификация лиц в реальном времени с использованием свёрточной нейронной сети и хэширующего леса
- Belkin, M. Laplacian eigenmaps and spectral techniques for embedding and clustering/M. Belkin, P. Niyogi//Proceedings of the 14th International Conference on Neural Information Processing Systems: Natural and Synthetic. -2001. -P. 585-591.
- Gionis, A. Similarity search in high dimensions via hashing/A. Gionis, P. Indyk, R. Motwani//Proceedings of the 25th International Conference on Very Large Data Bases. -1999. -P. 518-529.
- Gong, Y. Iterative quantization: A procrustean approach to learning binary codes for large-scale image retrieval/Y. Gong, S. Lazebnik, A. Gordo, F. Perronnin//IEEE Transactions on Pattern Analysis and Machine Intelligence. -2012. -Vol. 35, Issue 12. -P 2916-2929. - DOI: 10.1109/TPAMI.2012.193
- Grauman, K. Learning binary hash codes for large-scale image search/K. Grauman, R. Fergus. -In Book: Machine Learning for Computer Vision/ed. by R. Cipolla, S. Battiato, G.M. Farinella. -Berlin, Heidelberg: Springer, 2013. -P. 49-87. -ISBN: 978-3-642-28660-5. - DOI: 10.1007/978-3-642-28661-2_3
- He, K. K-means Hashing: An affinity-preserving quantization method for learning binary compact codes/K. He, F. Wen, J. Sun//IEEE Conference on Computer Vision and Pattern Recognition (CVPR). -2013. -P. 2938-2945. - DOI: 10.1109/CVPR.2013.378
- Irie, G. Locally linear hashing for extracting non-linear manifolds/G. Irie, L. Zhenguo, W. Xiao-Ming, C. Shih-Fu//IEEE Conference on Computer Vision and Pattern Recognition (CVPR). -2014. -P. 2115-2122. - DOI: 10.1109/CVPR.2014.272
- Liu, W. Supervised hashing with kernels/W. Liu, J. Wang, R. Ji, Y.-G. Jiang, S.-F. Chang//IEEE Conference on Computer Vision and Pattern Recognition (CVPR). -2012. -P. 2074-2081. - DOI: 10.1109/CVPR.2012.6247912
- Salakhutdinov, R. Semantic hashing/R. Salakhutdinov, G. Hinton//International Journal of Approximate Reasoning. -2009. -Vol. 50, Issue 7. -P. 969-978. - DOI: 10.1016/j.ijar.2008.11.006
- Shakhnarovich, G. Learning task-specific similarity: PhD thesis/G. Shakhnarovich. -Cambridge, MA: Massachusetts Institute of Technology, 2005.
- Shakhnarovich, G. Fast pose estimation with parameter sensitive hashing/G. Shakhnarovich, P. Viola, T. Darrell//Proceedings of the Ninth IEEE International Conference on Computer Vision. -2003. -Vol. 2. -P. 750-757. - DOI: 10.1109/ICCV.2003.1238424
- Weiss, Y. Spectral Hashing/Y. Weiss, A. Torralba, R. Fergus. -In Book: Advances in Neural Information Processing Systems 21 -Proceedings of the 2008 Conference. -2008. -P. 1753-1760.
- Zhang, L. Topology preserving hashing for similarity search/L. Zhang, Y. Zhang, X. Gu, J. Tang, Q. Tian//Proceedings of the 21st ACM international conference on Multimedia. -2013. -P. 123-132. - DOI: 10.1145/2502081.2502091
- Cao, Z. Face Recognition with Learning-based Descriptor/Z. Cao, Q. Yin, X. Tang, J. Sun//IEEE Conference on Computer Vision and Pattern Recognition (CVPR). -2010. -P. 2707-2714. - DOI: 10.1109/CVPR.2010.5539992
- Fan, H. Learning deep face representation/H. Fan, Z. Cao, Y. Jiang, Q. Yin, C. Doudou//arXiv preprint arXiv:1403.2802. -2014.
- Sun, Y. Deep learning face representation by joint identification-verification/Y. Sun, X. Wang, X. Tang//Proceedings of the 27th International Conference on Neural Information Processing Systems. -2014. -P. 1988-1996.
- Sun, Y. DeepID3: Face recognition with very deep neural networks/Y. Sun, X. Wang, X. Tang//arXiv preprint arXiv:1502.00873. -2015.
- Sun, Y. Deep learning face representation from predicting 10,000 classes/Y. Sun, X. Wang, X. Tang//Proceedings of the 2014 IEEE Conference on Computer Vision and Pattern Recognition. -2014. -P. 1891-1898. - DOI: 10.1109/CVPR.2014.244
- Taigman, Y. DeepFace: closing the gap to human-level performance in face verification/Y. Taigman, M. Yang, M. Ranzato, L. Wolf//Proceedings of the 2014 IEEE Conference on Computer Vision and Pattern Recognition. -2014. -P. 1701-1708. - DOI: 10.1109/CVPR.2014.220
- Wang, W. Face recognition based on deep learning/W. Wang, J. Yang, J. Xiao, S. Li, D. Zhou. -In Book: Human Centered Computing/ed. by Q. Zu, B. Hu, N. Gu, S. Seng. -Vol. 8944. -Springer, 2015. -P. 812-820. -ISBN: 978-3-319-15553-1. - DOI: 10.1007/978-3-319-15554-8_73
- Wu, X. Learning robust deep face representation/X. Wu//arXiv preprint arXiv:1507.04844. -2015.
- Zhou, E. Naive-deep face recognition: Touching the limit of LFW benchmark or not?/E. Zhou, Z. Cao, Q. Yin//arXiv preprint arXiv:1501.04690. -2015.
- Fan, H. Learning compact face representation: Packing a face into an int32/H. Fan, M. Yang, Z. Cao, Y. Jiang, Q. Yin//Proceedings of the 22nd ACM international conference on Multimedia. -2014. -P. 933-936. - DOI: 10.1145/2647868.2654960
- Chen, D. Blessing of dimensionality: High-dimensional feature and its efficient compression for face verification/D. Chen, X. Cao, F. Wen, J. Sun//Proceedings of IEEE Computer Society Conference on Computer Vision and Pattern Recognition. -2013. -P. 3025-3032. - DOI: 10.1109/CVPR.2013.389
- Nguyen, H.-V. Cosine similarity metric learning for face verification/H.-V. Nguyen, L. Bai//Proceedings of 10th Asian Conference on Computer Vision. -2010. -P. 709-720. - DOI: 10.1007/978-3-642-19309-5_55
- Taigman, Y. Multiple one-shots for utilizing class label information/Y. Taigman, L. Wolf, T. Hassner//Proceedings of British Machine Vision Conference. -2009. - DOI: 10.5244/C.23.77
- Liu, J. Targeting ultimate accuracy: face recognition via deep embedding/J. Liu, Y. Deng, T. Bai, Z. Wei, C. Huang//arXiv preprint arXiv:1506.07310. -2015.
- Qiu, Q. Random forests can hash/Q. Qiu, G. Sapiro, A. Bronstein//arXiv preprint arXiv:1412.5083. -2014.
- Vens, C. Random forest based feature induction/C. Vens, F. Costa//Proceedings of IEEE 11th International Conference on Data Mining (ICDM). -2011. -P. 744-753. - DOI: 10.1109/ICDM.2011.121
- Yu, G. Scalable forest hashing for fast similarity search/G. Yu, J. Yuan//Proceedings of IEEE International Conference on Multimedia and Expo (ICME). -2014. -P. 1-6. - DOI: 10.1109/ICME.2014.6890219
- Springer, J. Forest hashing: Expediting large scale image retrieval/J. Springer, X. Xin, Z. Li, J. Watt, A. Katsaggelos//Proceedings of IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). -2013. -P. 1681-1684. - DOI: 10.1109/ICASSP.2013.6637938
- Mishina, Y. Boosted random forest/Y. Mishina, R. Murata, M. Tsuchiya, H. Fujiyoshi//IEICE Transactions on Information and Systems. -2015. -Vol. E98-D, Issue 9. -P. 1630-1636. - DOI: 10.1587/transinf.2014OPP0004
- Best-Rowden, L. Unconstrained face recognition: Identifying a person of interest from a media collection/L. Best-Rowden, H. Han, C. Otto, B. Klare, A.K. Jain//IEEE Transactions on Information Forensics and Security. -2014. -Vol. 9, Issue 12. -P. 2144-2157. - DOI: 10.1109/TIFS.2014.2359577
- Huang, G.-B. Learning to align from scratch/G.-B. Huang, M. Mattar, H. Lee, E. Learned-Miller//Proceedings of the 25th International Conference on Neural Information Processing Systems. -2012. -P. 764-772.
- Schroff, F. FaceNet: A unified embedding for face recognition and clustering/F. Schroff, D. Kalenichenko, J. Philbin//Proceedings of IEEE Conference on Computer Vision and Pattern Recognition (CVPR). -2015. -P. 815-823. - DOI: 10.1109/CVPR.2015.7298682
