臺灣博碩士論文加值系統

在影像辨識的領域裡，從機器學習到神經網路的架構，再到卷積神經網路（Convolution Neural Networks，CNN），甚至發展出針對區域性的卷積神經網路，像是R-CNN、Fast R-CNN、Faster R-CNN等都是卷積神經網路發展的過程。
本論文使用Google公司開放的TensorFlow Object Detection API來訓練模型，此架構使用TensorFlow來實現大多數深度學習的目標檢測並用Python來撰寫程式。Python是最近興起的程式語言，其語法簡單且功能強大，再加上擁有數萬個套件可應用在各種領域上，使得在撰寫程式時能方便快速。由於影像辨識的發展，網路上已有許多圖像的數據庫供使用者使用並加以訓練成自己所要的辨識圖庫，本研究目的以針對貓和狗的品種加以分類並採用Faster R-CNN進行圖像識別，在圖庫中也可以增加自己所拍攝的圖片進行辨識。
在研究過程中，需要先安裝TensorFlow和Anaconda，之後就開始設定影像辨識的環境設定，最後就可以開始訓練模型，訓練完後就可以開始進行圖像的辨識。在訓練模型前需先將圖片進行標註並輸入至神經網路訓練，透過損失率觀察模型的收斂，訓練完後將神經網路輸出並儲存，圖像就可經系統的辨識得到座標並分類。

In the field of image recognition, from machine learning to the architecture of neural networks, to convolutional neural networks, and even the development of regional convolutional neural networks, such as R-CNN, Fast R-CNN, Faster R-CNN, etc. are all the development process of convolution neural network.
This paper uses Google's open TensorFlow Object Detection API to train the model. This architecture uses TensorFlow to achieve most deep learning target detection and write programs in Python. Python is a programming language that has emerged recently. Its syntax is simple and powerful. In addition, it has tens of thousands of packages that can be used in various fields, making it convenient and fast when writing programs. Due to the development of image recognition, there are many image databases on the Internet for users to use and train them into the recognition library they want. The purpose of this research is to classify cat and dog breeds and use Faster R-CNN for image Recognition, you can also add pictures you have taken for identification in the gallery.
In the research process, you need to install TensorFlow and Anaconda first, and then start to set the environment settings for image recognition. Finally, you can start training the model. After training, you can start image recognition. Before training the model, it is necessary to label the image and input it to the neural network for training. Observe the convergence of the model through the loss rate. After training, the neural network is output and stored, and the image can be identified and classified by the system.

摘要......i
Abstract......iii
誌謝......iv
目錄......v
表目錄......vii
圖目錄......viii
第一章 緒論......1
1.1研究動機......1
1.2研究目的......1
1.3論文架構......2
第二章 人工神經網路......3
2.1人工神經網路......3
2.1.1神經網路架構......3
2.1.2神經網路計算原理......4
2.1.3反向傳播......5
2.2卷積神經網路......6
2.2.1卷積層......6
2.2.2 池化層......7
2.2.3線性整流層......8
2.2.4全連結層......11
2.3卷積運算......12
2.3.1填充(Padding)......12
2.3.2滑步(stride)......13
2.3.3 3D卷積運算......14
第三章 Faster R-CNN與ResNet......15
3.1Faster R-CNN......15
3.1.1Faster R-CNN架構......16
3.1.2Region Proposal Network (RPN)......17
3.1.3 RoI pooling......19
3.1.4分類層......20
3.1.5訓練Faster R-CNN網路......20
3.2ResNet......25
3.2.1殘差學習......25
3.2.2ResNet101......26
3.2.3 基於ResNet101的Faster R-CNN......27
第四章 實驗結果與討論......28
4.1實驗架構......28
4.2軟體介紹......29
4.3實驗成果......30
第五章 結論與未來展望......51
5.1結論......51
5.2未來展望......51
參考文獻......52

[1]A. Krizhevsky, I. Sutskever, and G. E. Hinton, “Imagenet Classification with Deep Convolutional Neural Networks,” in Advances in neural information processing systems, pp. 1097–1105, 2012.
[2]J. Redmon, S. Divvala, R. Girshick, and A. Farhadi, “You Only Look Once: Unified, Real-Time Object Detection,” in Proceedings of the ieee conference on computer vision and pattern recognition, pp. 779–788, 2016.
[3]R. Girshick, “Fast R-Cnn,” in Proceedings of the ieee international conference on computer vision, pp. 1440–1448, 2015.
[4]S. Ren, K. He, R. Girshick, and J. Sun, “Faster R-Cnn: Towards Real-Time Object Detection with Region Proposal Networks,” in Advances in neural information processing systems 28, C. Cortes, N. D. Lawrence, D. D. Lee, M. Sugiyama, and R. Garnett, Eds. Curran Associates, Inc., pp. 91–99, 2015.
[5]J. Huang et al., “Speed/Accuracy Trade-Offs for Modern Convolutional Object Detectors,” in Proceedings of the ieee conference on computer vision and pattern recognition, pp. 7310–7311, 2017.
[6]Karen Simonyan, Andrew Zisserman, “Very Deep Convolutional Networks for Large-Scale Image Recognition” 3rd International Conference on Learning Representations, ICLR 2015, San Diego, CA, USA, May 7-9, 2015, Conference Track Proceedings. 2015
[7]C. Szegedy et al., "Going deeper with convolutions," 2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Boston, MA, pp. 1-9, 2015.
[8]K. He, X. Zhang, S. Ren and J. Sun, "Deep Residual Learning for Image Recognition," 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Las Vegas, NV, pp. 770-778, 2016.
[9]R. Girshick, J. Donahue, T. Darrell and J. Malik, "Rich Feature Hierarchies for Accurate Object Detection and Semantic Segmentation," 2014 IEEE Conference on Computer Vision and Pattern Recognition, Columbus, OH, pp. 580-587, 2014.
[10]“Artificial Neural Network.” https://en.wikipedia.org/wiki/Artificial_neural_network [Online accessed 2020-03-05].
[11]“類神經網路初探 基本架構與感知器”https://blog.birkhoff.me/introducing-artificial-
neural-network-1/ [Online accessed 2020-03-07].
[12]D. E. Rumelhart, G. E. Hinton, and R. J. Williams, “Learning Representations by Back-Propagating Errors,” Nature, vol. 323, pp. 533–536, Oct. 1986.
[13]“Gradient descent” https://en.wikipedia.org/wiki/Gradient_descent [Online accessed 2020-03-26].
[14]B. Liu, W. Zhao and Q. Sun, "Study of object detection based on Faster R-CNN," 2017 Chinese Automation Congress (CAC), Jinan, pp. 6233-6236, 2017.
[15]H. Zhu, Y. Qi, H. Shi, N. Li and H. Zhou, "Human Detection Under UAV: an Improved Faster R-CNN Approach," 2018 5th International Conference on Systems and Informatics (ICSAI), Nanjing, pp. 367-372, 2018.
[16]R. Kumar, M. Sharma, K. Dhawale and G. Singal, "Identification of Dog Breeds Using Deep Learning," 2019 IEEE 9th International Conference on Advanced Computing (IACC), Tiruchirappalli, India, pp. 193-198, 2019.
[17]“一文讀懂Faster RCNN” https://zhuanlan.zhihu.com/p/31426458 [Online accessed 2020-03-26].
[18]K. He, X. Zhang, S. Ren and J. Sun, "Spatial Pyramid Pooling in Deep Convolutional Networks for Visual Recognition," in IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 37, no. 9, pp. 1904-1916, 1 Sept. 2015.
[19]“ResNet解析” https://blog.csdn.net/lanran2/article/details/79057994 [Online accessed 2020-04-23].
[20]M. Abadi et al., “TensorFlow: Large-Scale Machine Learning on Heterogeneous Systems.” 2015.
[21]Q. Ye, X. Yang, C. Chen and J. Wang, "River Water Quality Parameters Prediction Method Based on LSTM-RNN Model," 2019 Chinese Control And Decision Conference (CCDC), Nanchang, China, pp. 3024-3028, 2019.
[22]林祐生，2019，”深度學習在機器人視覺辨識中的應用”，國立虎尾科技大學機械設計工程系研究所碩士論文。