請用此 Handle URI 來引用此文件：
http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/21591完整後設資料紀錄
| DC 欄位 | 值 | 語言 |
|---|---|---|
| dc.contributor.advisor | 傅楸善 | |
| dc.contributor.author | Tz-Chia Tseng | en |
| dc.contributor.author | 曾子家 | zh_TW |
| dc.date.accessioned | 2021-06-08T03:39:00Z | - |
| dc.date.copyright | 2019-07-26 | |
| dc.date.issued | 2019 | |
| dc.date.submitted | 2019-07-15 | |
| dc.identifier.citation | [1] A. Agarwal, et al., “Face anti-spoofing using haralick features,” In Biometrics Theory, Applications and Systems (BTAS), 2016 IEEE 8th International Conference, pp. 1-6, IEEE, 2016.
[2] Y. Agarwal, “Create your first Image Recognition Classifier using CNN, Keras and Tensorflow backend,” https://medium.com/nybles/create-your-first-image-recognition-classifier-using-cnn-keras-and-tensorflow-backend-6eaab98d14dd, 2019. [3] A. Anjos and S. Marcel, “Counter-measures to photo attacks in face recognition: A public database and a baseline,” In IJCB, 2011. [4] A. Anjos, et al., “Face anti-spoofing: Visual approach” In Handbook of Biometric Anti-Spoofing, pp. 65-82, Springer, 2014. [5] W. Bao, et al., “A liveness detection method for face recognition based on optical flow field,” In Image Analysis and Signal Processing, 2009, IASP 2009 International Conference, pp. 233-236, IEEE, 2009. [6] P. Bhandari, “Depth Estimation Techniques,” https://medium.com/@piyush1995bhandari/depth-estimation-techniques-830ffd297245, 2018. [7] Z. Boulkenafet, et al., “Face anti-spoofing based on color texture analysis,” CoRR, abs/1511.06316, 2015. [8] P. P. K. Chan, et al., “Face Liveness Detection Using a Flash against 2D Spoofing Attack,” IEEE Transactions on Information Forensics and Security, Volume 13, Issue 2, pp. 521–534, 2018. [9] I. Chingovska, et al., “On the effectiveness of local binary patterns in face anti-spoofing,” BIOSIG, pp. 1-7, 2012. [10] Common Objects in Context, http://cocodataset.org/#home, 2019. [11] E. D. Cubuk, et al., “AutoAugment: Learning Augmentation Policies from Data,” arXiv:1805.09501, 2018. [12] DeepLearning.Hub, “Learning non-maximum suppression,” https://twitter.com/DLdotHub/status/861966336527872000, 2019. [13] V. Dumoulin, F. Visin, “A guide to convolution arithmetic for deep learning,” arXiv:1603.07285, 2018. [14] N. Eineck, et al., “A multi-block-matching approach for stereo,” 2015 IEEE Intelligent Vehicles Symposium (IV), 2015. [15] W. Forson, “Understanding SSD MultiBox — Real-Time Object Detection In Deep Learning,” https://towardsdatascience.com/understanding-ssd-multibox-real-time-object-detection-in-deep-learning-495ef744fab, 2019. [16] H. Gao, et al., “Understand Single Shot MultiBox Detector (SSD) and Implement It in Pytorch,” https://medium.com/@smallfishbigsea/understand-ssd-and-implement-your-own-caa3232cd6ad, 2019. [17] R. GmbH, “Stereo matching,” https://doc.rc-visard.com/latest/en/stereo_matching.html#stereo-matching, 2018. [18] Google Open Source, “Open Images Dataset,” https://opensource.google.com/projects/open-images-dataset, 2019. [19] R. A. Hamzah1, and H. Ibrahim, “Literature Survey on Stereo Vision Disparity Map Algorithms,” https://www.hindawi.com/journals/js/2016/8742920/, 2016. [20] H. Hirschmuller, “Stereo Processing by Semi-Global Matching and Mutual Information,” IEEE transactions on pattern analysis and machine, 2007. [21] Kaggle , “How to choose CNN Architecture MNIST,” https://www.kaggle.com/cdeotte/how-to-choose-cnn-architecture-mnist, 2019. [22] R. Khandelwal, “Neural Network -Activation functions ,” https://medium.com/datadriveninvestor/neural-networks-activation-functions-e371202b56ff, 2019. [23] W. Kim, et al., “Face liveness detection from a single image via diffusion speed model,” IEEE Transactions on Image Processing, 24(8):2456-2465, 2015. [24] K. Kollreider, et al., “Evaluating liveness by face images and the structure tensor,” In IEEE Workshop on Automatic Identification Advanced Technologies, pp. 75-80, 2005. [25] N. Lakshminarayana, et al., “A discriminative spatio-temporal mapping of face for liveness detection,” 2017 IEEE International Conference on Identity, Security and Behavior Analysis (ISBA), 2017. [26] W. Liu, et al., “SSD: Single Shot MultiBox Detector,” arXiv:1512.02325, 2016. [27] J. Maatta, et al., “Face spoofing detection from single images using micro-texture analysis,” Biometrics (IJCB), 2011, international joint conference on, IEEE, pp. 1-7, 2011. [28] D. T. Nguyen, et al., “Combining Deep and Handcrafted Image Features for Presentation Attack Detection in Face Recognition Systems Using Visible-Light Camera Sensors,” arXiv:1804.06702, 2018. [29] OpenCV, “Depth Map from Stereo Images,” https://docs.opencv.org/3.1.0/dd/d53/tutorial_py_depthmap.html, 2019. [30] OpenCV, “Histogram Equalization,” https://opencv-python-tutroals.readthedocs.io/en/latest/py_tutorials/py_imgproc/py_histograms/py_histogram_equalization/py_histogram_equalization.html, 2019. [31] P. Pai, “Data Augmentation Techniques in CNN using Tensorflow,” https://medium.com/ymedialabs-innovation/data-augmentation-techniques-in-cnn-using-tensorflow-371ae43d5be9, 2019. [32] G. Pan, et al., “Eyeblink-based anti-spoofing in face recognition from a generic webcamera,” 2007 IEEE 11th International Conference on Computer Vision, 2007. [33] PASCAL2 Network, “The PASCAL Visual Object Classes Homepage,” http://host.robots.ox.ac.uk/pascal/VOC/, 2019. [34] T. F. Pereira, et al., “Face liveness detection using dynamic texture.,” EURASIP JIVP, 2014. [35] X. Qi, et al., “Comparison of Support Vector Machine and Softmax Classifiers in Computer Vision,” 2017 Second International Conference on Mechanical, Control and Computer Engineering, 2017. [36] J. Redmon, et al., “You Only Look Once: Unified, Real-Time Object Detection,” arXiv:1506.02640, 2019. [37] S. Ren, et al., “Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks,” arXiv:1506.01497, 2015. [38] X. Song, et al., “Discriminative Representation Combinations for Accurate Face Spoofing Detection,” arXiv:1808.08802, 2018. [39] Stanford University, “CS231n Convolutional Neural Networks for Visual Recognition,” http://cs231n.github.io/neural-networks-1/, 2019. [40] D. Tang, et al., “Face Flashing: a Secure Liveness Detection Protocol Based on Light Reflections,” arXiv:1801.01949, 2018. [41] R. Tronci, et al., “Fusion of multiple clues for photo-attack detection in face recognition systems,” In 2011 IEEE International Joint Conference on Biometrics, IJCB 2011, Washington, DC, USA, pp. 1-6, 2011. [42] S. T. Tsang, “Review: VGGNet — 1st Runner-Up (Image Classification), Winner (Localization) in ILSVRC 2014,” https://medium.com/coinmonks/paper-review-of-vggnet-1st-runner-up-of-ilsvlc-2014-image-classification-d02355543a11, 2018. [43] S. -H. Tsang, “Review: SSD — Single Shot Detector (Object Detection),” https://towardsdatascience.com/review-ssd-single-shot-detector-object-detection-851a94607d11, 2019. [44] UFLDL Tutorial , “Convolutional Neural Network,” http://deeplearning.stanford.edu/tutorial/supervised/ConvolutionalNeuralNetwork/, 2019. [45] J. Li, Y. Wang, et al., “Live face detection based on the analysis of Fourier spectra,” In Biometric Technology for Human Identification, pp. 296-303, 2004. [46] Wikipedia, ”Adaptive histogram equalization,” https://en.wikipedia.org/wiki/Adaptive_histogram_equalization, 2019. [47] Wikipedia, “Field-programmable gate array,” https://en.wikipedia.org/wiki/Field-programmable_gate_array, 2019. [48] Wikipedia, “Hamming distance,” https://en.wikipedia.org/wiki/Hamming_distance, 2019. | |
| dc.identifier.uri | http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/21591 | - |
| dc.description.abstract | 本論文提出一個在不藉助深度相機鏡頭、紅外線熱感應鏡頭等額外裝置，僅使用智慧型手機單顆可見光前鏡頭，辨識取得的影像來自真人使用者，或電子欺騙攻擊 (Spoofing Attack)，包含紙本相片、電子相片、影片等。首先我們透過單樣本多框偵測器 (Single Shot Multi-Box Detector, SSD) 擷取畫面中的人臉，經過特定的資料擴增 (Data Augmentation) 後，再餵入捲積類神經網路 (Convolution Neural Network, CNN) ，訓練出具有分辨真人使用者與電子欺騙攻擊的模型。 | zh_TW |
| dc.description.abstract | Our proposed method is capable of authenticating the input image is from real user or spoofing attack, including paper photograph, digital photograph, and video, using only the Red, Green, Blue (RGB) frontal camera of common smart phone, without the help of depth camera or infrared thermal sensor. We first capture live faces in each frame of input video streams by single shot multi-box detector then feed into our designed convolution neural network after certain data augmentation and finally obtain a well-trained spoof face classifier. | en |
| dc.description.provenance | Made available in DSpace on 2021-06-08T03:39:00Z (GMT). No. of bitstreams: 1 ntu-108-R05945052-1.pdf: 3825847 bytes, checksum: d11275252f81bd4ceedca01eff97d694 (MD5) Previous issue date: 2019 | en |
| dc.description.tableofcontents | CONTENTS
誌謝 i 中文摘要 iii ABSTRACT iv CONTENTS v LIST OF FIGURES ix LIST OF TABLES xii Chapter 1 Introduction 1 1.1 Overview 1 1.2 Acquisition of Training Images 3 1.3 Thesis Organization 4 Chapter 2 Related Work 5 2.1 Overview 5 2.2 Stereo Matching for Disparity and Depth 8 2.2.1 Local Methods 13 2.2.2 Global Methods 15 2.2.3 Semi-Global Methods 16 2.2.4 The Road Not Taken 16 Chapter 3 Background 17 3.1 Convolutional Neural Network 17 3.1.1 Architecture 18 3.1.2 Modeling one neuron 21 3.1.3 Strides 23 3.1.4 Pooling Layer 24 3.1.5 Convolution Layer 26 3.1.6 Fully Connected Layer 29 3.1.7 Activation Function 31 3.1.8 Softmax 33 3.2 Data Augmentation 34 3.2.1 Scaling 35 3.2.2 Translation 36 3.2.3 Rotation (at 90 Degrees): 36 3.2.4 Rotation (at Finer Angles): 36 3.2.5 Flipping: 38 3.2.6 Adding Salt and Pepper Noise: 39 3.2.7 Lighting Condition: 40 3.2.8 Perspective Transform: 40 Chapter 4 Methodology 42 4.1 Overview 42 4.2 Single Shot Multi-Box Detector (SSD) 43 4.2.1 Training 47 .4.2.1.1 Generate Priors 47 .4.2.1.2 MultiBox Detector 47 .4.2.1.3 Match Priors with Ground-Truth Boxes 48 .4.2.1.4 Scale Ground-Truth Boxes 50 .4.2.1.5 Hard Negative Mining 51 .4.2.1.6 The Loss Function 52 .4.2.1.7 Data Augmentation 52 4.2.2 Prediction 53 .4.2.2.1 NMS 54 4.2.3 The Drawbacks 56 4.3 Histogram Equalization 56 4.3.1 Histogram 56 4.3.2 Histogram Equalization 57 4.3.3 Adaptive Histogram Equalization 58 4.3.4 Contrastive Limited Adaptive Equalization 60 4.4 Data Augmentation 62 4.5 Neural Network 64 4.5.1 VGG-16 64 4.5.2 The architecture of neural network 66 Chapter 5 Experimental Result 72 5.1 Overview 72 5.2 Evaluation 73 5.3 Results and Observations 75 Chapter 6 Conclusion and Future Work 79 References 80 LIST OF FIGURES Figure 2 1: RGB, depth map from depth camera and depth map from stereo matching. [6] 10 Figure 2 2: The notations and their relationship of stereo matching. [29] 12 Figure 3 1: Basic CNN structure. [21] 18 Figure 3 2: First layer of a convolutional neural network with pooling. Units of the same color have tied weights and units of different color represent different filter maps. [44] 20 Figure 3 3: Neuron unit. [39] 22 Figure 3 4: Stride of 2 pixels. [25] 23 Figure 3 5: An image max-pooled with 2x2 filters and stride 2. [25] 25 Figure 3 6: Convolution arithmetic. [13] 27 Figure 3 7: A 32x32 image padded with 2 pixels on each side. [13] 28 Figure 3 8: Some common filters. [25] 29 Figure 3 9: After pooling layer, flattened as FC layer. [25] 30 Figure 3 10: Complete CNN architecture. [25] 31 Figure 3 11: The pattern of a ReLU function. [22] 32 Figure 3 12: The pipeline of a softmax function. 33 Figure 3 13: Notice the background noise added in the image set. [31] 38 Figure 3 14: Salt and pepper noise added. [31] 39 Figure 3 15: Perspective transform. [31] 41 Figure 4 1: Pipeline of our proposed method. 42 Figure 4 2: VGG-based SSD architecture. [16] 46 Figure 4 3: SSD: Multiple Bounding Boxes for Localization (loc) and Confidence (conf). [43] 48 Figure 4 4: Diagram explaining IoU. [15] 49 Figure 4 5: Pseudo-code of Jaccard index. [16] 50 Figure 4 6: Example of hard negative mining. [15] 51 Figure 4 7: Some “Crazy” Detection Results on MS COCO Dataset. [15] 54 Figure 4 8: Pseudo code of NMS. 55 Figure 4 9: NMS example. [12] 55 Figure 4 10: The initial image and its histogram and cumulative histogram. [30] 57 Figure 4 11: The initial after histogram equalization. [30] 58 Figure 4 12: The comparison of initial image, histogram equalized image, and adaptive histogram equalized image. [30] 59 Figure 4 13: The initial image, histogram equalized image, and contrastive limited adaptive equalized image. [Sudhakar, 2019] 61 Figure 4 14: The initial image and contrastive limited adaptive equalized image. 62 Figure 4 15: Different VGG Layer Structures Using Single Scale (256) Evaluation. [42] 65 Figure 4 16: The architecture of our proposed neural network. 71 Figure 5 1: The training loss and accuracy on dataset. 73 Figure 5 2: Snapshots of testing results in true positive group. 75 Figure 5 4: Snapshots of testing results in true negative group. 76 Figure 5 5: Snapshots of testing results in miss detection group. 77 Figure 5 6: Snapshots of testing results in false alarm group. 78 LIST OF TABLES Table 5 1: The environment to develop and conduct experiments is shown in this table. 72 Table 5 2: The testing result: percentage (number of frames / number of total frames). 74 | |
| dc.language.iso | en | |
| dc.title | 智慧型手機真人臉型驗證 | zh_TW |
| dc.title | Anti-Spoofing of Live Face Authentication on Smartphone | en |
| dc.type | Thesis | |
| dc.date.schoolyear | 107-2 | |
| dc.description.degree | 碩士 | |
| dc.contributor.oralexamcommittee | 江元麟,鄭宇哲,王獻章 | |
| dc.subject.keyword | 活臉辨識,電子欺騙攻擊,單樣本多框偵測器,資料擴增,捲積類神經網路, | zh_TW |
| dc.subject.keyword | Live Face Authentication,Spoofing Attack,Single Shot Multi-Box Detector,Data Augmentation,CNN, | en |
| dc.relation.page | 86 | |
| dc.identifier.doi | 10.6342/NTU201901370 | |
| dc.rights.note | 未授權 | |
| dc.date.accepted | 2019-07-15 | |
| dc.contributor.author-college | 電機資訊學院 | zh_TW |
| dc.contributor.author-dept | 生醫電子與資訊學研究所 | zh_TW |
| 顯示於系所單位： | 生醫電子與資訊學研究所 | |
文件中的檔案：
| 檔案 | 大小 | 格式 | |
|---|---|---|---|
| ntu-108-1.pdf 未授權公開取用 | 3.74 MB | Adobe PDF |
系統中的文件，除了特別指名其著作權條款之外，均受到著作權保護，並且保留所有的權利。