臺灣博碩士論文加值系統

卷積神經網絡（CNN）是一種前饋神經網絡，他由若干捲積層和池化層組成，尤其在圖像分類任務中表現出很高的效率。在這項研究中，採用CNN多標籤分類從服裝圖像中提取特徵並對服裝類型和顏色進行分類。本篇論文所使用的方法為，先分別為類型和顏色開發神經網絡，然後組合成一個多輸出模型來識別它們。然後，我們利用準確性、精確度、召回率和f1分數來評估我們的網絡。在最後實驗結果證明，該方法在分類精度方面達到了實用性。

Convolutional Neural Network (CNN) has demonstrated great efficiency in image classification tasks. In this research, CNN is employed with multi-label classification to extract features from clothing images and classify clothing types and colors. In our method, we first develop a neural network for types and colors separately then combine into a multi-output model to identify them. Then, we evaluate our network by accuracy, precision, recall and f1-score. The experimental results show that the proposed method achieves practical performance in classify precision.

Chinese Abstract ………………………………………………………….………… i
English Abstract ………………………………………………………….………… ii
Acknowledgements ………………………………………………………….……….... iii
Table of Contents ………………………………………………………….………… iv
List of Tables ………………………………………………………….………… v
List of Figures ………………………………………………………….………… vi
List of Abbreviations ………………………………………………………….………… vii
Chapter 1 Introduction……………………………………………………… 1
1.1 Research Motivation……………………………………………... 1
1.2 Thesis Objectives ………………………………………………... 2
1.3 Thesis Outline……..……………………………………………... 2
Chapter 2 Background and Related Work ….………….…………………… 3
2.1 Convolutional Neural Network ………………………………….. 3
2.1.1 Inputs and Outputs ...…………………………………………….. 3
2.1.2 Convolutional layer...…………………………………………….. 4
2.1.3 Pooling layer.…………………………………………………….. 5
2.1.4 Fully connected layer …...……………………………………….. 6
2.2 Multi-output Classification Model……………………………….. 6
2.3 Data augmentation.…………………………...………………….. 7
2.4 Performance metric.…………………………..………………….. 8
2.4.1 Confusion Matrix…..…………………………………………….. 8
2.4.2 Accuracy…………...…………………………………………….. 8
2.4.3 Precision…………....…………………………………………….. 9
2.4.4 Recall…………….....…………………………………………….. 9
2.4.5 F1 Score…………....…………………………………………….. 9
2.5 Related Work……..…………………………..………………….. 10
2.5.1 Clothing type classification.…………………..………………….. 10
2.5.2 Color classification.…………………….……..………………….. 10
2.5.3 Multi-label classification….…………………..………………….. 11
Chapter 3 Methodology…………………….……………………………….. 12
3.1 Method……..…………………………………………………….. 12
3.1.1 Data Collection…………....…………………..………………….. 12
3.1.2 Data Preparation…………..…………………..………………….. 13
3.1.3 Training Model..…………..…………………..………………….. 14
3.1.3.1 Category classification………..…………………..……………… 15
3.1.3.2 Color classification………..…………………..………………….. 15
3.1.3.3 Category and color classification………..…………………..…… 16
3.2 Material.………………………………………………………….. 17
3.2.1 R………………………………………………………………….. 17
3.2.2 Keras…….……………………………………………………….. 18
Chapter 4 Findings.………………………………………….………………. 19
4.1 Dataset...………………………………………………………….. 19
4.2 Experiment……………………………………………………….. 21
4.2.1 Experiment 1.…………………………………………………….. 21
4.2.2 Experiment 2.…………………………………………………….. 22
4.3 Evaluation.……………………………………………………….. 23
Chapter 5 Conclusion………………………………...................…………… 25
Reference ………………...…………………………………….……………. 26
Extended Abstract ………………...…………………………………….……………. 28

[1] Chaitawat Chenbunyanon, Ji-Han Jiang. Clothing classification with multi-attribute using convolutional neural network, International Computer Symposium(ICS), 2018.
[2] Juergen Schmidhuber. Deep learning in neural networks: An overview, Neural Networks vol. 61, pp. 85-117, (2015).
[3] Agnieszka Mikołajczyk, Michał Grochowski. Data augmentation for improving deep learning in image classification problem, International Interdisciplinary Ph.D. Workshop (IIPhDW), pp. 117-122, (2018).
[4] L. Perez, J. Wang. The Effectiveness of Data Augmentation in Image Classification using Deep Learning, arXiv preprint arXiv:1712.04621, (2017).
[5] Li Fengxin, Li Yueping, Zhang Xiaofeng. A novel approach to cloth classification through deep neural networks, International Conference on Security, Pattern Analysis, and Cybernetics (SPAC), pp. 368 – 371, (2017)
[6] Kimitoshi Yamazaki, Masayuki Inaba. Clothing classification using image features derived from clothing fabrics, wrinkles and cloth overlaps, IEEE/RSJ International Conference on Intelligent Robots and Systems, pp. 2710 – 2717, (2013).
[7] Reza Fuad Rachmadi, I Ketut Eddy Purnama. Vehicle Color Recognition using Convolutional Neural Network, arXiv preprint arXiv:1510.07391, (2015).
[8] Shobhit Bhatnagar, Deepanway Ghosal, Maheshkumar H. Kolekar. Classification of fashion article images using convolutional neural networks, Fourth International Conference on Image Information Processing (ICIIP), pp. 357-362, (2017).
[9] Abdallah Zeggada, Farid Melgani. Multilabel classification of UAV images with Convolutional Neural Networks, IEEE International Geoscience and Remote Sensing Symposium (IGARSS), pp. 5083 – 5086, (2016).
[10] Yunchao Wei, Wei Xia, Junshi Huang, Bingbing Ni, Jian Dong, Yao Zhao, Shuicheng Yan. CNN: Single-label to Multi-label, arXiv:1406.5726v3, (2014).
[11]Peng-Fei Zhang, Hao-Yi Wu, Xin-Shun Xu. A Dual-CNN Model for Multi-label Classification by Leveraging Co-occurrence Dependencies Between Labels, Advances in Multimedia Information Processing – PCM 2017, pp. 315-324, (2017).
[12]Keras. https://keras.io/, last accessed 2018/09/30.
[13]Keras functional API.https://keras.io/getting-started/functional-api-guide/, last accessed 2018/09/30
[14]K. Simonyan and A. Zisserman. Very deep convolutional networks for large-scale image recognition. Computer Science, arXiv:1409.1556v6, (2014)
[15]Hidenori Ide, Takio Kurita. Improvement of learning for CNN with ReLU activation by sparse regularization, International Joint Conference on Neural Networks (IJCNN), pp. 2684- 2691, (2017).
[16]Yunchao Gong, Yangqing Jia, Thomas Leung, Alexander Toshev, Sergey Ioffe. Deep Convolutional Ranking for Multilabel Image Annotation, arXiv preprint arXiv:1312.4894v2, (2013).
[17]R . https://www.r-project.org/, last accessed 2019/01/08.