跳到主要內容

臺灣博碩士論文加值系統

(98.84.25.165) 您好!臺灣時間:2024/11/10 00:04
字體大小: 字級放大   字級縮小   預設字形  
回查詢結果 :::

詳目顯示

: 
twitterline
研究生:陳鵬安
論文名稱:基於銳化卷積神經網路之人臉身份辨識
論文名稱(外文):Face Recognition based on Sharpening Convolutional Neural Network
指導教授:陳文雄陳文雄引用關係
口試委員:林祐仲王仁澤
口試日期:2017-07-20
學位類別:碩士
校院名稱:國立暨南國際大學
系所名稱:電機工程學系
學門:工程學門
學類:電資工程學類
論文種類:學術論文
論文出版年:2017
畢業學年度:105
語文別:中文
論文頁數:45
中文關鍵詞:工業4.0人臉辨識大數據深度學習Sharpening收斂速度過擬合一般化
外文關鍵詞:Industrial 4.0face recognitionBig Datadeep learningSharpeningtraining convergence rateoverfittinggeneralization
相關次數:
  • 被引用被引用:0
  • 點閱點閱:242
  • 評分評分:
  • 下載下載:23
  • 收藏至我的研究室書目清單書目收藏:1
隨著資訊科技的蓬勃發展,由德國發起工業4.0(第四次工業革命)也即將到來,大數據因此成為主流,身份辨識系統也受到相當的衝擊。然而傳統人臉識別方法多半使用手工選擇特徵方法,這種方法造成在學習大數據多層模型時候,會造成參數過多或是學習特徵不夠好,因此像深度學習這種需要大數據庫學習模型因此成為主流,深度學習可以把利用機器自動學習任務目標的特徵,也因此深度學習這種架構成為學術非常熱門的技術。
這種多層深度學習架構有著許多種模型,其中本論文使用的是卷積神經網路,會使用卷積神經網路是因為影像是二維方式,卷積神經網路就把神經網路當成二維影像神經網路,並且保持圖像中位置之間的關係在神經網路中,並且利用局部連接與權重共享分式讓我們網路參數不會過於龐大。
神經網路中有可能因為激活函數選擇或是學習率大小導致訓練困難。本論文提出一種Sharpening方法去強化卷積過後特徵圖,在Yale資料庫上面測試,發現使我們訓練與測試時候收斂速度有明顯提升,經過比較多實驗比較過後,我們 發現最後發現此方法在測試階段有可能會訓練過度影響到測試性能結果,所以未來改善方向可以加入一些神經網路中防止過擬合提升一般化方法進行更好改良。

Recently, deep neural network has achieved a promising result in face recognition. Neural networks have different models in which convolutional deep neural network (DCNN) is a powerful visual model that yields a hierarchical feature map structure. However, in convolutional neural network too many super-parameters result in training difficulties. Therefore, vanishing gradient or exploding gradient problems are an eternal topic in deep learning.
Thesis proposes a sharpening convolutional neural network to solve this problem. This model can execute multi-class face identification tasks, since sharpening convolutional neural network greatly improves the model generation capacity by introducing an effective sharpening layer. This new architectural model consists of sharpening layers which can emphasize or enhance the feature maps, so that it may improve the training convergence rate. The experimental results show that embedding sharpening layers in DCNN may extract better features.

目次
誌謝辭 i
摘要 ii
Abstract iii
目次 iv
表目次 vi
圖目次 vii
第一章 緒論 1
1.1前言 1
1.2人臉識別 2
1.2.1 人臉檢測 3
1.2.2 特徵提取 4
1.2.3 分類器 5
1.3研究動機 7
1.4論文架構 8
第二章 深度學習簡介 9
2.1深度學習的核心概念 11
2.2深度學習的架構 11
2.2.1限制波茲曼機(Restricted Boltzmann Machine) 11
2.2.2深信度網路(Deep Belief Networks) 11
2.2.3 卷積神經網路(Convolutional Neural Networks) 12
2.2.4 自動編碼(Auto Encoder) 15
2.2.5 稀疏編碼(Sparse Coding) 15
2.3 卷積網路人臉應用參考文獻回顧 17
2.4 卷積網路架構改動參考文獻回顧 18
第三章 卷積神經網路實作方法 20
3.1數據預處理 21
3.1.1簡單縮放 21
3.1.2逐樣本均值削減 21
3.1.3特徵標準化 21
3.2銳化卷積神經網路架構 22
3.3網路深度探討 23
3.4卷積核尺寸大小(Convolutional Kernel Size) 24
3.5激活函數選擇 25
第四章 實驗結果與分析 27
4.1實驗平台 27
4.2實驗資料庫 27
4.3交叉驗證 28
4.4從神經元來分析Sharpening 29
4.5結果與分析 34
第五章 結論與未來展望 42
5.1結論 42
5.2未來展望 42
參考文獻 43


參考文獻

[1] Hinton, Geoffrey E, and Ruslan R. Salakhutdinov. "Reducing the dimensionality of data with neural networks." science, vol. 313, no. 5786 , pp. 504-507, July 2006.
[2] Hinton, Geoffrey E, Simon Osindero, and Yee-Whye Teh. "A fast learning algorithm for deep belief nets." Neural computation, vol. 18, no. 7, pp. 1527-1554, 2006.
[3] LeCun, Yann, et al. "Gradient-based learning applied to document recognition." Proceedings of the IEEE, vol. 86, no. 11, pp. 2278-2324, 1998.
[4] Krizhevsky, Alex, Ilya Sutskever, and Geoffrey E. Hinton. "Imagenet classification with deep convolutional neural networks." Advances in neural information processing systems, 2012.
[5] Boureau, Y-Lan, et al. "Learning mid-level features for recognition." Computer Vision and Pattern Recognition (CVPR), 2010 IEEE Conference on. IEEE, 2010.
[6] Boureau, Y-Lan, Jean Ponce, and Yann LeCun. "A theoretical analysis of feature pooling in visual recognition." Proceedings of the 27th international conference on machine learning (ICML-10), 2010.
[7] Bengio, Yoshua, et al. "Greedy layer-wise training of deep networks." Advances in neural information processing systems, 2007.
[8] Olshausen, Bruno A, and David J. Field. "Emergence of simple-cell receptive field properties by learning a sparse code for natural images." Nature, vol. 381, no. 6583, pp. 607, 1996.
[9] Sun, Yi, Xiaogang Wang, and Xiaoou Tang. "Hybrid deep learning for face verification." Proceedings of the IEEE International Conference on Computer Vision, 2013.
[10] Sun, Yi, Xiaogang Wang, and Xiaoou Tang. "Deep learning face representation from predicting 10,000 classes." Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2014.
[11] Schroff, Florian, Dmitry Kalenichenko, and James Philbin. "Facenet: A unified embedding for face recognition and clustering." Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2015.
[12] Zhu, Zhenyao, et al. "Recover canonical-view faces in the wild with deep neural networks." arXiv preprint arXiv, vol. 1404, no. 3543, 2014.
[13] Jaderberg, Max, Karen Simonyan, and Andrew Zisserman. "Spatial transformer networks." Advances in Neural Information Processing Systems, 2015.
[14] Springenberg, Jost Tobias, et al. "Striving for simplicity: The all convolutional net." arXiv preprint arXiv, vol. 1412, no. 6806, 2014.
[15] Lin, Min, Qiang Chen, and Shuicheng Yan. "Network in network." arXiv preprint arXiv, vol. 1312, no. 4400, 2013.
[16] Zhao, Haishan, et al. "STAT3 genetic variant, alone and in combination with STAT5b polymorphism, contributes to breast cancer risk and clinical outcomes." Medical Oncology, vol. 32, no. 1, pp. 375, 2015.
[17] Yang, Zichao, et al. "Deep fried convnets." Proceedings of the IEEE International Conference on Computer Vision, 2015.
[18] Jeng, Ren-He. "多模態生物辨識: 方法與應用." 暨南大學電機工程學系學位論文, pp. 1-54, 2016.
[19] Hinton, Geoffrey E., and Terrence J. Sejnowski. "Learning0 and releaming in Boltzmann machines." Parallel Distrilmted Processing , vol. 1, 1986.
[20] http://cs.nyu.edu/~fergus/tutorials/deep_learning_cvpr12/tutorial_p2_nnets_
ranzato_short.pdf.
[21] http://deeplearning.stanford.edu/wiki/index.php/UFLDL_Tutorial.
[22] https://forum.stanford.edu/events/2011/2011slides/plenary/2011plenaryNg.pdf.
[23] http://cs.nyu.edu/~fergus/tutorials/deep_learning_cvpr12/CVPR2012-Tutorial_lee.pdf.

QRCODE
 
 
 
 
 
                                                                                                                                                                                                                                                                                                                                                                                                               
第一頁 上一頁 下一頁 最後一頁 top
無相關期刊