在電腦視覺領域中，人臉偵測已經是一個行之有年的研究領域，從Viola-Jones 提出的Haar 特徵值偵測一直到近年深度學習網路(Deep Learning)，電腦偵測人臉的能力一直不斷進步，其偵測率甚至在近年已相當接近實際人眼觀察能力，然而從一些特別困難的人臉資料庫如WIDER-Faces中，我們還是可以看得出有許多的改進空間，特別是當人臉相當微小，具體而言當圖像中的人臉區域像素值長或寬不超過10~20的時候，儘管人眼還是多少可以觀測出來，但電腦視覺演算法普遍還是難以偵測這些細小人臉。
近年來有非常多表現特別出色的人臉偵測的深度卷積式深層網路(Deep-CNN)，也不斷嘗試解決細小人臉偵測不佳的問題，其普遍的作法分為兩種 - 利用影像金字塔或深度網路特徵金字塔提升偵測率，我們會於論文中概述兩者優點與缺點，而本篇論文同樣是針對極度細小人臉的偵測能力加強，而我們則訴求在不額外增加影像或特徵金字塔的前提下，配合現有之人臉偵測技術的資訊，使用近年來同樣蓬勃發展的超解析度化(Super resolution)技術提升並圖像中細小人臉區域，在充足的圖像資訊條件下，分析並進一步去提升人臉偵測率結果。
我們所提出的方法理論上可應用於現有的許多主流人臉偵測演算法，而實驗結果將會提供近年來表現優異的深層網路進行深入的資料庫分析比對，最後透過視覺化圖，證實我們的方法確實有發揮增強細小人臉偵測率的作用。

Object detection has been a hot research topic for many years to date. Among various task we have achieved so far, recent state-of-the-art face detection through deep learning architecture method have proven to be very robust to many kinds of image scenario. Yet when it comes to very tiny faces, typically faces that has no more than 10~20 in pixel sizes, still remains a challenging task for face detection.
Many methods have been proposed to address tiny face detection problem. Most of them are based on image-pyramid or feature-pyramid method. By leveraging multiple sizes in spatial domain or receptive field in convolutional feature layers, it’s a common believe that we can extract more information and thus getting a better result in face detection. Both of these method have their pros and cons.
In this paper, we aim our goal to the tiny face detection and introduce a method that utilize super resolution technique which can increase the sample candidate from the existing detection pipeline and enhance the performance without additional image or feature pyramids. We provided detailed comparison and visualization in our work, and the result have shown performance boost in face detection task.

摘 要 i
Abstract ii
Content iii
Chapter 1 Introduction 4
1.1 Research Background 4
1.2 Motivation 5
1.3 Contribution 7
Chapter 2 Related Work 8
2.1 Object and Face Detection 8
2.2 Super Resolution and Face Hallucination 9
2.3 Generative Adversarial Network 11
Chapter 3 Proposed Method 12
3.1 Overview of Network 12
3.2 Detection Pipeline 13
3.3 Proposed Network for Tiny Face SR 14
3.4 Post-processing of Bounding Boxes Information 18
Chapter 4 Experiments and Discussions 19
4.1 Premise of Baseline Comparison 19
4.2 Training Details and Performance Evaluations 20
4.3 Observations and Measurement Difficulties 25
4.4 Measurement Corrected Result 27
Chapter 5 Conclusion 30
References 31

[1] S. Yang, P. Luo, C. C. Loy, and X. Tang, “WIDER FACE: A face detection benchmark,” in Proc. IEEE Conf. Comput. Vis. Pattern Recognit., Las Vegas, NV, USA, Jun. 2016, pp. 5525–5533.
[2] K. Zhang, Z. Zhang, Z. Li, and Y. Qiao, “Joint face detection and alignment using multitask cascaded convolutional networks,” IEEE Signal Process. Lett., vol. 23, no. 10, pp. 1499–1503, Oct. 2016.
[3] C. Zhu, Y. Zheng, K. Luu, and M. Savvides, “CMS-RCNN: Contextual multi-scale region-based cnn for unconstrained face detection,” arXiv:1606.05413.
[4] P. F. Felzenszwalb, R. B. Girshick, D. McAllester, and D. Ramanan, “Object detection with discriminatively trained part-based models,” IEEE Trans. Pattern Anal. Mach. Intell., vol. 32, no. 9, pp. 1627–1645, Sept. 2010.
[5] S. Yang, Y. Xiong, C. C. Loy, and X. Tang, “Face detection through scale-friendly deep convolutional networks,” arXiv:1706.02863.
[6] T. Lin, P. Dollár, R. Girshick, K. He, B. Hariharan, and S. Belongie, “Feature pyramid networks for object detection,” in Proc. IEEE Conf. Comput. Vis. Pattern Recognit., Honolulu, HI, USA, Jul. 2017. pp. 2117–2125.
[7] P. Hu, and D. Ramanan, “Finding tiny faces,” in Proc. IEEE Conf. Comput. Vis. Pattern Recognit., Honolulu, HI, USA, Jul. 2017. pp. 951–959.
[8] Z. Liu, P. Luo, X. Wang, and X. Tang, “Deep learning face attributes in the wild,” in Proc. IEEE Int. Conf. Comput. Vis., Dec. 2015, pp. 3730–3738.
[9] S. Zhu, S. Liu, C. C. Loy, and X. Tang, “Deep cascaded bi-network for face hallucination,” in Proc. Eur. Conf. Comput. Vis., Oct. 2016. pp. 614–630.
[10] H. Bin, C. Weihai, W. Xingming, and L. Chun-Liang, “High-quality face image sr using conditional generative adversarial networks,” arXiv:1707.00737.
[11] R. Dahl, M. Norouzi, and J. Shlens, “Pixel recursive super resolution,” in Proc. IEEE Int. Conf. Comput. Vis., Oct. 2017, pp. 5439–5448.
[12] S. Ren, K. He, R. Girshick, and J. Sun, “Faster r-cnn: Towards real-time object detection with region proposal networks,” IEEE Trans. Pattern Anal. Mach. Intell., vol. 39, no. 6, pp. 1137–1149, Jun. 2017.
[13] A. Shrivastava, A. Gupta, and R. Girshick, “Training region-based object detectors with online hard example mining,” in Proc. IEEE Conf. Comput. Vis. Pattern Recognit., Las Vegas, NV, USA, Jun. 2016, pp. 761–769.
[14] E. Shelhamer, J. Long, and T. Darrell, “Fully convolutional networks for semantic segmentation,” IEEE Trans. Pattern Anal. Mach. Intell., vol. 39, no. 4, pp. 640–651, Apr. 2017.
[15] K. Simonyan, and A. Zisserman, “Very deep convolutional networks for large-scale image recognition,” in Proc. Int. Conf. Learn. Representations., 2015
[16] K. He, X. Zhang, S. Ren, and J. Sun, “Deep residual learning for image recognition,” in Proc. IEEE Conf. Comput. Vis. Pattern Recognit., Las Vegas, NV, USA, Jun. 2016, pp. 770–778.
[17] C. Dong, C. C. Loy, K. He and X. Tang, “Image super-resolution using deep convolutional networks,” IEEE Trans. Pattern Anal. Mach. Intell., vol. 38, no. 2, pp. 295–307, Feb. 2016.
[18] C. Dong, C. C. Loy, and X. Tang, “Accelerating the super-resolution convolutional neural network,” in Proc. Eur. Conf. Comput. Vis., Oct. 2016, pp. 391–407.
[19] J. Kim, J. K. Lee, and K. M. Lee, “Accurate image super-resolution using very deep convolutional networks,” in Proc. IEEE Conf. Comput. Vis. Pattern Recognit., Las Vegas, NV, USA, Jun. 2016, pp. 1646–1654.
[20] W. S. Lai, J. B. Huang, N, Ahuja, and M. H. Yang, “Deep laplacian pyramid networks for fast and accurate super-resolution,” in Proc. IEEE Conf. Comput. Vis. Pattern Recognit., Honolulu, HI, USA, Jul. 2017. pp. 624–632.
[21] J. Wu, S. Ding, W. Xu, and H. Chao, “Deep joint face hallucination and recognition,” arXiv:1611.08091, 2016
[22] I. Goodfellow, J. Pouget-Abadie, M. Mirza, B. Xu, D. Warde-Farley, S. Ozair, A. Courville, and Y. Bengio, “Generative adversarial nets,” in Proc. Adv. Neural Inf. Process. Syst., 2014, pp. 2672–2680.
[23] M. Arjovsky, S. Chintala, and L. Bottou, “Wasserstein gan,” arXiv:1701.07875, 2017.
[24] I. Gulrajani, F. Ahmed, M. Arjovsky, V. Dumoulin, and A. Courville, “Improved training of wasserstein gans,” arXiv:1704.00028, 2017.
[25] C. Ledig, L. Theis, F. Huszar, J. Caballero, A. Aitken, A. Tejani, J. Totz, Z. Wang, and W. Shi, “Photo-realistic single image super-resolution using a generative adversarial network,” in Proc. IEEE Conf. Comput. Vis. Pattern Recognit., Honolulu, HI, USA, Jul. 2017. pp. 4681–4690.
[26] X. Yu, and F. Porikli, “Ultra-resolving face images by discriminative generative networks,” in Proc. Eur. Conf. Comput. Vis., Oct. 2016, pp. 318–333.