人工智慧是近幾年來一直很熱門的一個領域，其中的機器學習裡的深度學習已廣泛應用在語音辨識、文本生成、影像辨識等生活中常見的例子裡。本研究使用深度學習中的捲積神經網路對鋼片製成中的鋼片品質進行監測，論文中建立3種捲積網路模型3Conv、4Conv及5Conv，來對鋼片進行缺障判斷。論文中我們對各個模型進行分析與比較，實驗結果顯示其準確率(accuracy)分別是：3Conv為79.75%、4Conv為77.50%及5Conv為74.99%，而其損失函數(loss function)值分別是：3Conv為0.4108、4Conv為0.5957及5Conv為0.6736。所以由實驗結果數據可知3Conv表現最佳，可達到將近8成的準確率。

　　Artificial intelligence is an area that has been popular in recent years, among them, deep learning in machine learning has been widely used in common examples in life such as speech recognition, text generation, and image recognition. This research uses a convolutional neural network in deep learning to monitor the quality of steel sheets in the production of steel sheets, Three convolutional network models 3Conv, 4Conv and 5Conv are established in the paper to judge the defect of the steel sheet. In the paper, we analyze and compare each model, The experimental results show that their accuracy rates are: 3Conv 79.75%, 4Conv 77.50% and 5Conv 74.99%, The loss function values are: 3Conv 0.4108, 4Conv 0.5957 and 5Conv 0.6736. So from the experimental data, we know that 3Conv performs best, Can achieve nearly 80% accuracy.

誌謝　i
摘要　ii
Abstract　iii
目錄　iv
圖目錄　vii
表目錄　ix

第一章　緒論　1
　1.1 研究背景與動機　1
　1.2 研究目的　2
第二章　文獻探討　3
　2.1 人工智慧(Artificial Intelligence, AI)　3
　2.2 機器學習(Machine Learning)　4
　　2.2.1 監督式學習(Supervised Learning)　4
　　2.2.2 非監督式學習(Unsupervised Learning)　5
　2.3 類神經網路(Artificial Neural Network, ANN)　5
　　2.3.1 生物神經網路　5
　　2.3.2 神經元　8
　　2.3.3 激活函數(Activation function)　9
　　2.3.4 類神經網路之架構　13
　2.4 捲積神經網路(Convolutional Neural Network, CNN)　16
　　2.4.1 捲積層(Convolutional Layer)　18
　　2.4.2 池化層(Pooling Layer)　20
　　2.4.3 全連接層(Fully connected Layer)　21
　　2.4.4 AlexNet模型　22
　　2.4.5 VGG模型　23
第三章　研究方法　26
　3.1 研究架構　27
　3.2 資料準備　28
　3.3 建立模型架構　31
　　3.3.1 6層網路模型架構(3Conv)　31
　　3.3.2 7層網路模型架構(4Conv)　34
　　3.3.3 8層網路模型架構(5Conv)　37
　3.4 訓練模型　40
　3.5 評估模型　41
第四章　實驗結果　42
　4.1 資料擴增法(Data Augmentation)　42
　4.2 模型訓練與測試結果　43
　　4.2.1 6層網路模型(3Conv)　44
　　4.2.2 7層網路模型(4Conv)　45
　　4.2.3 8層網路模型(5Conv)　46
　4.3 評估模型　47
第五章　結論　49
參考文獻　50

李漢銘(2004)。生物神經元模型。取自 http://neuron.csie.ntust.edu.tw/homework/93/nn/homework2/m9304302/12.htm
斎藤康毅(2017)。Deep Learning：用Python進行深度學習的基礎理論實作。台北市：基峰資訊。
Chapelle, O., Schölkopf, B., & Zien, A. (2006). Semi-Supervised Learning. IEEE Transactions on Neural Networks, 20(3), 542-542.
Combinido, J. S., Mendoza, J. R., & Aborot, J. (2018). A Convolutional Neural
cnblogs(2016)。淺談深度學習中的激活函數 - The Activation Function in Deep
Francois Chollet(2019)。Deep Learning深度學習必讀 - Keras大神帶你用Python實作。臺北市：旗標。
Frossard, D. (2016). Model and pre-trained parameters for VGG16 in TensorFlow. From https://www.cs.toronto.edu/~frossard/post/vgg16/
GCP專門家(2018)。人工智慧、機器學習、深度學習是什麼? -Machine Learning教學系列。取自https://blog.gcp.expert/ml-1-ai-ml-deep-learning-intro/
Hastie, T., Tibshirani, R., & Friedman, J. (2009). Unsupervised learning. In The elements of statistical learning. New York, NY, Springer.
Jitender_1998. Implement sigmoid function using Numpy. From https://www.geeksforgeeks.org/implement-sigmoid-function-using-numpy/
Krizhevsky, A., Sutskever, I., & Hinton, G. E. (2017). ImageNet Classification with Deep Convolutional Neural Networks. ACM, 60(6), 84-90.
Lecun, Y., Bottou, L., Bengio, Y., & Haffner, P. (1998). Gradient-based learning applied to document recognition. Proceedings of the IEEE, 86(11), 2278-2324.
Learning.取自https://read01.com/JNzJkL.html#.XrqYz0QzbIV
Ng, A. (2018). What is a deep neural network. From https://www.itread01.com/content/1543043179.html
Network Approach for Estimating Tropical Cyclone Intensity Using Satellite-based Infrared Images. IEEE ICPR, Beijing, China.
Simonyan, K., & Zisserman, A. (2015). Very Deep Convolutional Networks For Large-S Cale Image Recognition. ICLR, arXiv: 1409.1556.
Sultan, H. H., Salem, N. M., & AL-Atabany, W. (2019). Multi-classification of Brain Tumor Images using Deep Neural Network. IEEE Access, 7(1), 69215-69225.
Tommy Huang(2018)。卷積神經網路(Convolutional neural network, CNN) - CNN運算流程。取自https://medium.com/@chih.sheng.huang821/%E5%8D%B7%E7%A9%8D%E7%A5%9E%E7%B6%93%E7%B6%B2%E8%B7%AF-convolutional-neural-network-cnn-cnn%E9%81%8B%E7%AE%97%E6%B5%81%E7%A8%8B-ecaec240a631
Tommy Huang(2018)。機器學習-神經網路(多層感知機 Multilayer perceptron, MLP)運作方式。取自https://medium.com/@chih.sheng.huang821/%E6%A9%9F%E5%99%A8%E5%AD%B8%E7%BF%92-%E7%A5%9E%E7%B6%93%E7%B6%B2%E8%B7%AF-%E5%A4%9A%E5%B1%A4%E6%84%9F%E7%9F%A5%E6%A9%9F-multilayer-perceptron-mlp-%E9%81%8B%E4%BD%9C%E6%96%B9%E5%BC%8F-f0e108e8b9af