臺灣博碩士論文加值系統

近年來，由於深度學習演算法的突飛猛進、高效硬體平台不斷推陳出新、及網路上存在的多樣性影像資料，使得以深度學習演算法為基礎的圖像辨識精準度大幅提高，電腦視覺相關應用也隨之蓬勃發展。然而深度學習的推論結果與領域知識及訓練測試資料存在密切的關係，唯有持續利用最新資料不斷訓練深度學習模型，方可配合環境變更與時俱進，維持模型辨識的高精準度。
本研究使用Nvidia GPU以軟體貨櫃技術（Docker Container）為基礎，搭配Nvidia Docker及配置Docker Compose建置視覺物件偵測（Object Detection）與訓練平台，並結合持續整合及交付（Continuous Integration / Continuous Delivery）持續利用誤告警（False Alarm）資訊透過重新標記自動再訓練模型，維持模型一定的準確度。
軟體貨櫃技術的高彈性及輕量化特性，實現快速替換視覺物件偵測框架（Framework），結合本研究實作視覺物件偵測的各組元件，除了快速替換不同框架，更可依據不同視覺物件偵測的輸入、輸出、演算法等進行快速替換，實現系統各項分工，以利深度學習之應用實際落地。

In recent years, due to the rapid development of deep learning algorithms, the continuous development of high-efficiency hardware platforms, and the diversity of image data on the Internet, the accuracy of image recognition based on deep learning algorithms has been greatly improved, and computer vision-related applications have also been improved. It will flourish. However, the inference results of deep learning are closely related to domain knowledge and training test data. Only by continuously using the latest data to continuously train the deep learning model can we keep pace with the times and maintain the high accuracy of model identification.
This study uses Nvidia GPUs based on Docker Containers, with Nvidia Docker and Docker Compose to build visual object detection and training platforms, combined with continuous integration and continuous delivery (Continuous Integration / Continuous Delivery) continuous use of false alarms (False Alarm) information automatically retrains the model by re-marking to maintain a certain degree of accuracy.
The high flexibility and light weight of the soft container technology enables the rapid replacement of the visual object detection framework (Framework), combined with the components of the visual object detection in this study, in addition to quickly replacing different frames, it can also be based on different visual objects. The measured input, output, and algorithm are quickly replaced to realize the division of labor of the system, so as to facilitate the application of deep learning.

中文摘要 i
Abstract ii
謝誌 iii
目 錄 iv
表目錄 v
圖目錄 vi
第壹章緒論 8
第一節研究背景 8
第二節 研究動機與目的 9
第三節 章節介紹 9
第貳章文獻探討 10
第一節 智慧警監 10
第二節 影像辨識 13
第三節 BigData 16
第四節 CI/CD 18
第五節 Docker 20
第參章系統架構與設計 22
第一節 系統架構 22
第二節 系統運作流程 24
第肆章系統實作 26
第一節 設計系統Dockerfile 28
第二節 物件偵測系統實作 29
第三節 系統雛型展示 33
第伍章結論與未來研究方向 38
第一節結論 38
第二節 未來研究方向 38
參考文獻 39
博碩士論文紙本授權書 42

[1]Abadi, M., Barham, P., Chen, J., Chen, Z., Davis, A., Dean, J., ... & Kudlur, M. (2016). Tensorflow: A system for large-scale machine learning. In 12th {USENIX} Symposium on Operating Systems Design and Implementation ({OSDI} 16) (pp. 265-283).
[2]Apache MXNet, Retrieved May 31, 2019, from https://mxnet.apache.org/
[3]Azevedo, A. I. R. L., & Santos, M. F. (2008). KDD, SEMMA and CRISP-DM: a parallel overview. IADS-DM.
[4]Cattell, R. (2011). Scalable SQL and NoSQL data stores. Acm Sigmod Record, 39(4), 12-27.
[5]Chen, L. (2015). Continuous delivery: Huge benefits, but challenges too. IEEE Software, 32(2), 50-54.
[6]Chen, T., Li, M., Li, Y., Lin, M., Wang, N., Wang, M., ... & Zhang, Z. (2015). Mxnet: A flexible and efficient machine learning library for heterogeneous distributed systems. arXiv preprint arXiv:1512.01274.
[7]Docker, Retrieved May 31, 2019, from https://www.docker.com/
[8]Fowler, M., & Foemmel, M. (2006). Continuous integration. Thought-Works) http://www. thoughtworks. com/Continuous Integration. pdf, 122, 14.
[9]Ghemawat, S., Gobioff, H., & Leung, S. T. (2003). The Google file system.
[10]Girshick, R. (2015). Fast r-cnn. In Proceedings of the IEEE international conference on computer vision (pp. 1440-1448).
[11]Girshick, R., Donahue, J., Darrell, T., & Malik, J. (2014). Rich feature hierarchies for accurate object detection and semantic segmentation. In Proceedings of the IEEE conference on computer vision and pattern recognition (pp. 580-587).
[12]Grafana, Retrieved May 31, 2019, from https://grafana.com/
[13]InfluxDB, Retrieved May 31, 2019, from https://www.influxdata.com/
[14]Katal, A., Wazid, M., & Goudar, R. H. (2013). Big data: Issues, challenges, tools and Good practices. 2013 Sixth International Conference on Contemporary Computing (IC3). doi:10.1109/ic3.2013.6612229
[15]Krizhevsky, A., Sutskever, I., & Hinton, G. E. (2012). Imagenet classification with deep convolutional neural networks. In Advances in neural information processing systems (pp. 1097-1105).
[16]LeCun, Y., Bengio, Y., & Hinton, G. (2015). Deep learning. Nature, 521(7553), 436–444. doi:10.1038/nature14539
[17]Mikolov, T., Karafiát, M., Burget, L., Černocký, J., & Khudanpur, S. (2010). Recurrent neural network based language model. In Eleventh annual conference of the international speech communication association.
[18]Mitchell, T., Buchanan, B., DeJong, G., Dietterich, T., Rosenbloom, P., & Waibel, A. (1990). Machine Learning. Annual Review of Computer Science, 4(1), 417–433. doi:10.1146/annurev.cs.04.060190.002221
[19]Nvidia Docker, Retrieved May 31, 2019, from https://github.com/NVIDIA/nvidia-docker
[20]Paszke, A., Gross, S., Chintala, S., Chanan, G., Yang, E., DeVito, Z., ... & Lerer, A. (2017). Automatic differentiation in pytorch.
[21]Pytorch, Retrieved May 31, 2019, from https://pytorch.org/
[22]Ren, S., He, K., Girshick, R., & Sun, J. (2015). Faster r-cnn: Towards real-time object detection with region proposal networks. In Advances in neural information processing systems (pp. 91-99).
[23]Sharif Razavian, A., Azizpour, H., Sullivan, J., & Carlsson, S. (2014). CNN features off-the-shelf: an astounding baseline for recognition. In Proceedings of the IEEE conference on computer vision and pattern recognition workshops (pp. 806-813).
[24]Shvachko, K., Kuang, H., Radia, S., & Chansler, R. (2010, May). The hadoop distributed file system. In MSST (Vol. 10, pp. 1-10).
[25]Tensorflow, Retrieved May 31, 2019, from https://www.tensorflow.org/
[26]Turing, A. M. (2009). Computing machinery and intelligence. In Parsing the Turing Test (pp. 23-65). Springer, Dordrecht.
[27]Vavilapalli, V. K., Murthy, A. C., Douglas, C., Agarwal, S., Konar, M., Evans, R., ... & Saha, B. (2013, October). Apache hadoop yarn: Yet another resource negotiator. In Proceedings of the 4th annual Symposium on Cloud Computing (p. 5). ACM.
[28]Where does AlphaGo go: from church-turing thesis to AlphaGo thesis and beyond. (2016). IEEE/CAA Journal of Automatica Sinica, 3(2), 113–120. doi:10.1109/jas.2016.7471613
[29]Wirth, R., & Hipp, J. (2000, April). CRISP-DM: Towards a standard process model for data mining. In Proceedings of the 4th international conference on the practical applications of knowledge discovery and data mining (pp. 29-39). Citeseer.
[30]張淂福(2019年09月25日) 。談安防監控的大數據與智慧影像分析技術。全球安防科技網。取自https://www.asmag.com.tw/showpost/10837.aspx