在過去的幾十年裡，人臉識別領域的研究取得了很大的進展。然而，當受試者臉部具有偽裝，比如眼鏡，化妝和帽子等等，臉部驗證的問題就是一個具有挑戰性的問題。
現有的偽裝人臉識別的方法使用了經典的人臉檢測器。但是，當人臉數量增加時，檢測時間明顯增長。此外，對於有側臉的情況，偵測到的特徵點的位置不准確。
在本文中，我們不是直接使用經典的人臉檢測器，而是提出了一種基於BiSeNet的用於人臉解析的方法，這個方法將圖像轉換為標籤圖。通過這種表示，我們得到了臉部特徵像素級的座標，實現了可以通過逐像素標籤圖來表示臉部。
對於臉部驗證，我們使用了深度卷積神經網路的網路架 ResNet-101，用於識別穿著偽裝的人和冒名頂替者。此外，結合了由標籤圖給出的臉部屬性的豐富位置信息，我們最終在最近推出Disguised Faces in the Wild資料集中展示了結果，並證明了基於BiSeNet的人臉解析方法和深度卷積神經網路的方法，對於區分偽裝面和冒名頂替者是有效的；另外我們的網絡在化妝資料集中的表現優於目前最好的方法。

Research in the field of face recognition has made great progress over the past couple of decades. However, the problem of face verification is a challenging problem when the subjects wearing a disguise like glasses, makeup and hat, etc..
The existing disguised face recognition methods with classic face detector. But when the number of faces increases, the detection time rise obviously. Also, the position of the feature points is not accurate for face in profile.

In this paper, instead of using the classic face detector directly, we propose a BiSeNet based approach for face parsing, which transformed an image patch to a label map. With this representation, we get the pixel-level coordinates of facial features, a face can also be presented by pixel-wise label maps. For face verification, we use ResNet-101 which is a deep convolutional neural network for recognizing people wearing disguises and the impostors. Also, rich location information of facial attributes given by the label map. We finally present results on the recently introduced Disguised Faces in the Wild challenge dataset and show that the method via DCNN and BiSeNet are effective for discriminating between disguised faces and impostors in the wild. In addition, our network outperforms state-of-the-art results on makeup face verification in makeup datasets.

[1] A. Kumar, A. A., and Chellappa, R. Kepler: keypoint and pose estimation of unconstrained faces by learning efficient h-cnn regressors. 258-265.
[2] AmarjotSingh,DevendraPatil,G.M.R.,andOmkar,S.Disguisedfaceidenti- fication (dfi) with facial keypoints using spatial fusion convolutional network. In IEEE International Conference on Computer Vision Workshops (ICCVW) (2017).
[3] Bansal, A., Nanduri, A., Castillo, C. D., Ranjan, R., and Chellappa, R. Umd- faces: An annotated face dataset for training deep networks. arXiv preprint arXiv:1611.01484v2 (2016).
[4] Bulat, A., and Tzimiropoulos, G. How far are we from solving the 2d & 3d face alignment problem? (and a dataset of 230,000 3d facial landmarks). In International Conference on Computer Vision (2017).
[5] C. Szegedy, S. I., and Vanhoucke, V. Inception-v4, inception-resnet and the impact of residual connections on learning. In ICLR Workshop (2017).
[6] Chen, D., Hua, G., Wen, F., and Sun, J. Supervised transformer network for efficient face detection. CoRR abs/1607.05477 (2016).
[7] Deng, J., Guo, J., Niannan, X., and Zafeiriou, S. Arcface: Additive angular margin loss for deep face recognition. In CVPR (2019).
[8] F. Zhou, J. B., and Lin, Z. Exemplar-based graph matching for robust facial landmark localization. In IEEE International Conference on Computer Vision (ICCV) (2013).
[9] Florian Schroff, Dmitry Kalenichenko, J. P. Facenet: A unified embedding for face recognition and clustering. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (2015).
[10] Guo, G., Wen, L., and Yan, S. Face authentication with makeup changes. IEEE Transactions on Circuits and Systems for Video Technology 24 (2014), 814–825.
[11] Hu, J., Ge, Y., Lu, J., and Feng, X. Makeup-robust face verification. 2013 IEEE International Conference on Acoustics, Speech and Signal Processing (2013), 2342–2346.
[12] Huang, G. B., Ramesh, M., Berg, T., and Learned-Miller, E. Labeled faces in the wild: A database for studying face recognition in unconstrained envi- ronments. Tech. Rep. 07-49, University of Massachusetts, Amherst, October 2007.
[13] J. Hu, J. L., and Tan, Y.-P. Discriminative deep metric learning for face ver- ification in the wild. In IEEE Conference on Computer Vision and Pattern Recognition (2014), pp. 1875–1882.
[14] J. Hu, L. S., and Sun, G. Squeeze-and-excitation networks. arXiv. 32

[15] Jiang, H., and Learned-Miller, E. Face detection with the faster r-cnn. 650- 657.
[16] K. He, X. Zhang, S. R., and Sun, J. Deep residual learning for image recogni- tion. 770-778.
[17] Kaipeng Zhang, Y.-L. C., and Hsu, W. Deep disguised faces recognition. In
IEEE Conference on Computer Vision and Pattern Recognition (CVPR) 2018 - Disguised Faces in the Wild Workshop (2018).
[18] Krähenbühl, P., and Koltun, V. Efficient inference in fully connected crfs with gaussian edge potentials. In NIPS (2011).
[19] L. Liang, F. Wen, X. T., and Xu, Y. An integrated model for accurate shape alignment. In ECCV (2006).
[20] L. Liang, F. Wen, Y. X. X. T., and Shum, H. Accurate face alignment using shape constrained markov network. In CVPR (2006).
[21] Lee, C.-H., Liu, Z., Wu, L., and Luo, P. Maskgan: Towards diverse and inter- active facial image manipulation. Technical Report (2019).
[22] Li, Y., Song, L., Wu, X., He, R., and Tan, T. Anti-makeup: Learning A bi-level adversarial network for makeup-invariant face verification. CoRR abs/1709.03654 (2017).
[23] Liu, Z., Luo, P., Wang, X., and Tang, X. Deep learning face attributes in the wild. In Proceedings of International Conference on Computer Vision (ICCV) (December 2015).
[24] Luo, P., W. X. T. X. Hierarchical face parsing via deep learning. 2480–2487.
[25] M. Valstar, B. Martinez, X. B., and Pantic, M. Facial point detection using
boosted regression and graph models. In CVPR (2010).
[26] O. M. Parkhi, A. Vedaldi, A. Z. e. a. Deep face recognition. 6.
[27] Peri, S. V., and Dhall, A. Disguisenet: A contrastive approach for disguised face verification in the wild. In CVPR Workshop on Disguised Faces in the Wild (2018), vol. 4.
[28] R. Ranjan, C. D. C., and Chellappa, R. L2-constrained softmax loss for dis- criminative face verification. arXiv.
[29] Ranjan, R., Sankaranarayanan, S., Castillo, C. D., and Chellappa, R. An all- in-one convolutional neural network for face analysis. CoRR abs/1611.00851 (2016).
[30] Ronneberger, O., P.Fischer, and Brox, T. U-net: Convolutional networks for biomedical image segmentation. In Medical Image Computing and Computer- Assisted Intervention (MICCAI) (2015), vol. 9351 of LNCS, Springer, pp. 234– 241. (available on arXiv:1505.04597 [cs.CV]).
33

[31] S. Sankaranarayanan, A. Alavi, C. D. C., and Chellappa, R. Triplet probabilis- tic embedding for face verification and clustering. 1–8.
[32] SifeiLiu,Yang,J.,ChangHuang,andYang,M.Multi-objectiveconvolutional learning for face labeling. In 2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) (June 2015), pp. 3451–3459.
[33] Simonyan, K., and Zisserman, A. Very deep convolutional networks for large- scale image recognition. arXiv.
[34] Sun, Y., Ren, L., Wei, Z., Liu, B., Zhai, Y., and Liu, S. A weakly supervised method for makeup-invariant face verification. Pattern Recogn. 66, C (June 2017), 153–159.
[35] Tero Karras, Timo Aila, S. L. J. L. Progressive growing of gans for improved quality, stability, and variation.
[36] Tran, L., Yin, X., and Liu, X. Disentangled representation learning gan for pose-invariant face recognition. In In Proceeding of IEEE Computer Vision and Pattern Recognition (Honolulu, HI, July 2017).
[37] V. Kushwaha, M. Singh, R. S. M. V. N. R., and Chellappa, R. Disguised faces in the wild. technical report, iiit delhi.
[38] V. Le, J. Brandt, L. B. Z. L., and Huang, T. Interactive facial feature localiza- tion. In European Conference Computer Vision (ECCV) (2012).
[39] Y. Guo, L. Zhang, Y. H. X. H., and Gao, J. A dataset and benchmark for large-scale face recognition. 87–102.
[40] Y. Zhou, W. Zhang, X. T., and Shum, H. A bayesian mixture model for multi- view face alignment. In CVPR (2005).
[41] Yu, C., Wang, J., Peng, C., Gao, C., Yu, G., and Sang, N. Bisenet: Bilat- eral segmentation network for real-time semantic segmentation. In European Conference on Computer Vision (2018), Springer, pp. 334–349.
[42] Zhang, K., Zhang, Z., Li, Z., and Qiao, Y. Joint face detection and alignment using multitask cascaded convolutional networks. IEEE Signal Processing Letters 23, 10 (Oct 2016), 1499–1503.
[43] Zitnick, C. L., and Dollar, P. Edge boxes: Locating object proposals from edges. 391-405.