Real-time face identification via CNN and boosted hashing forest
Yu.V. Vizilter, V.S. Gorbatsevich, A.V. Vorotnikov, N.A. Kostromov

 

State Research Institute of Aviation Systems (GosNIIAS)

Full text of article: Russian language.

 PDF

Abstract:
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.

Keywords:
convolutional neural networks, hashing, binary trees, Hamming distance, biometrics.

Citation:
Vizilter YuV, Gorbatsevich VS, Vorotnikov AV, Kostromov NA. Real-time face identification via CNN and boosted hashing forest. Computer Optics 2017; 41(2): 254-265. DOI: 10.18287/2412-6179-2017-41-2-254-265.

References:

  1. Belkin M, Niyogi P. Laplacian eigenmaps and spectral techniques for embedding and clustering. Proc NIPS 2001; 14: 585-591.
  2. Gionis A, Indyk P, Motwani R. Similarity search in high dimensions via hashing. Proc VLDB 1999: 518-529.
  3. Gong Y, Lazebnik S, Gordo A, Perronnin F. Iterative quantization: A procrustean approach to learning binary codes for large-scale image retrieval. IEEE Trans Pattern Anal Mach Intell 2012; 35(12): 2916-2929. DOI: 10.1109/TPAMI.2012.193.
  4. Grauman K, Fergus R. Learning binary hash codes for large-scale image search. In: Cipolla R, Battiato S, Farinella GM, eds. Machine Learning for Computer Vision. Berlin, Heidelberg: Springer; 2013: 49-87. ISBN: 978-3-642-28660-5. DOI: 10.1007/978-3-642-28661-2_3.
  5. He K, Wen F, Sun J. K-means Hashing: An affinity-preserving quantization method for learning binary compact codes. Proc CVPR 2013: 2938-2945. DOI: 10.1109/CVPR.2013.378.
  6. Irie G, Zhenguo L, Xiao-Ming W, Shih-Fu C. Locally linear hashing for extracting non-linear manifolds. Proc CVPR 2014: 2115-2122. DOI: 10.1109/CVPR.2014.272.
  7. Liu W, Wang J, Ji R, Jiang Y-G, Chang S-F. Supervised hashing with kernels. Proc CVPR 2012: 2074-2081.  DOI: 10.1109/CVPR.2012.6247912.
  8. Salakhutdinov R, Hinton G. Semantic hashing. International Journal of Approximate Reasoning 2009; 50(7): 969-978. DOI: 10.1016/j.ijar.2008.11.006.
  9. Shakhnarovich G. Learning task-specific similarity. PhD thesis. Cambridge, MA: Massachusetts Institute of Technology; 2005.
  10. Shakhnarovich G, Viola P, Darrell T. Fast pose estimation with parameter sensitive hashing, Proc ICCV '03 2003; 2: 750-757. DOI: 10.1109/ICCV.2003.1238424.
  11. Weiss Y, Torralba A, Fergus R. Spectral Hashing. In Book: Advances in Neural Information Processing Systems 21 – Proceedings of the 2008 Conference 2008: 1753-1760.
  12. Zhang L, Zhang Y, Gu X, Tang J, Tian Q. Topology preserving hashing for similarity search. Proc ACM Int Conf Multimedia 2013: 123-132.  DOI: 10.1145/2502081.2502091.
  13. Cao Z, Yin Q, Tang X, Sun J. Face Recognition with Learning-based Descriptor. Proc CVPR 2010: 2707-2714.  DOI: 10.1109/CVPR.2010.5539992.
  14. Fan H, Cao Z, Jiang Y, Yin Q, Doudou C. Learning deep face representation. arXiv preprint arXiv:1403.2802 2014.
  15. Sun Y, Wang X, Tang X. Deep learning face representation by joint identification-verification. Proc NIPS 27 2014: 1988-1996.
  16. Sun Y, Wang X, Tang X. DeepID3: Face recognition with very deep neural networks. arXiv preprint arXiv:1502.00873 2015.
  17. Sun Y, Wang X, Tang X. Deep learning face representation from predicting 10,000 classes. Proc CVPR 2014: 1891-1898.  DOI: 10.1109/CVPR.2014.244.
  18. Taigman Y, Yang M, Ranzato M, Wolf L. DeepFace: closing the gap to human-level performance in face verification. Proc CVPR 2014: 1701-1708. DOI: 10.1109/CVPR.2014.220.
  19. Wang W, Yang J, Xiao J, Li S, Zhou D. Face recognition based on deep learning. In Book: Zu Q, Hu B, Gu N, Seng S, eds. Human Centered Computing. Springer; 2015: 812-820. ISBN: 978-3-319-15553-1. DOI: 10.1007/978-3-319-15554-8_73.
  20. Wu X. Learning robust deep face representation. arXiv preprint arXiv:1507.04844 2015.
  21. Zhou E, Cao Z, Yin Q. Naive-deep face recognition: Touching the limit of LFW benchmark or not? arXiv preprint arXiv:1501.04690 2015.
  22. Fan H, Yang M, Cao Z, Jiang Y, Yin Q. Learning Compact Face Representation: Packing a Face into an int32. Proc ACM Int Conf Multimedia 2014: 933-936. DOI: 10.1145/2647868.2654960.
  23. Chen D, Cao X, Wen F, Sun J. Blessing of dimensionality: High-dimensional feature and its efficient compression for face verification. Proc CVPR 2013: 3025-3032. DOI: 10.1109/CVPR.2013.389.
  24. Nguyen H-V, Bai L. Cosine similarity metric learning for face verification. Proc ACCV 2010: 709-720. DOI: 10.1007/978-3-642-19309-5_55.
  25. Taigman Y, Wolf L, Hassner T. Multiple one-shots for utilizing class label information. Proc BMVC 2009. DOI: 10.5244/C.23.77.
  26. Liu J, Deng Y, Bai T, Wei Z, Huang C. Targeting ultimate accuracy: face recognition via deep embedding. arXiv preprint arXiv:1506.07310 2015.
  27. Qiu Q, Sapiro G, Bronstein A. Random forests can hash, arXiv preprint arXiv:1412.5083 2014.
  28. Vens C, Costa F. Random Forest Based Feature Induction. Proc ICDM 2011: 744-753.  DOI: 10.1109/ICDM.2011.121.
  29. Yu G, Yuan J. Scalable forest hashing for fast similarity search. Proc ICME 2014: 1-6. DOI: 10.1109/ICME.2014.6890219.
  30. Springer J, Xin X, Li Z, Watt J, Katsaggelos A. Forest hashing: Expediting large scale image retrieval. Proc ICASSP 2013: 1681-1684. DOI: 10.1109/ICASSP.2013.6637938.
  31. Mishina Y, Murata R, Tsuchiya M. Fujiyoshi H. Boosted Random Forest. IEICE Transactions on Information and Systems 2015; E98-D(9): 1630-1636. DOI: 10.1587/transinf.2014OPP0004.
  32. Best-Rowden L, Han H, Otto C, Klare B, Jain AK. Unconstrained face recognition: Identifying a person of interest from a media collection. IEEE Trans Inf Forens Security 2014; 9(12): 2144-2157. DOI: 10.1109/TIFS.2014.2359577.
  33. Huang G-B, Mattar M, Lee H, Learned-Miller E. Learning to align from scratch. Proc NIPS '12 2012: 764-772.
  34. Schroff, F. Kalenichenko D. and Philbin J. FaceNet: A unified embedding for face recognition and clustering. Proc CVPR 2015: 815-823.  DOI: 10.1109/CVPR.2015.7298682.
© 2009, IPSI RAS
Institution of Russian Academy of Sciences, Image Processing Systems Institute of RAS, Russia, 443001, Samara, Molodogvardeyskaya Street 151; E-mail: journal@computeroptics.ru; Phones: +7 (846) 332-56-22, Fax: +7 (846) 332-56-20