卷積神經網路的架構是有多層的設計，在其中運作流程相當複雜，在架構最佳化的過程難以觀察調整參數所造成的影響。本論文以YOLOv2物件識別演算法為研究主軸，以圖像視覺化的方式呈現其運作流程。從中使用了JavaScript Library的D3.js，因為它可支援SVG格式及融合網頁開發的兩項特性。本論文開發的使用者互動系統，包含三大模組: YOLOv2架構特徵圖、權重圖展示模組，卷積展示模組，最後輸出張量錨框參數展示系統。將複雜的YOLOv2物件辨識過程，利用簡單的動態圖像呈現，使用者可以透過本系統網頁互動之方式瞭解複雜的演算法。此外，調整頁面上的參數時，圖像會即時更新，視覺化的互動可以幫助剛入門的使用者提升學習興趣。同時，以圖像表達可以加快對演算法架構的理解。對於有相關學習經驗的使用者，可以透過視覺化系統幫助自己學習。瞭解在使用物件識別技術所發生的問題，減少花費時間在錯誤的理解上。本系統讓學習者用較容易的方式，入門物件辨識的相關知識。

The architecture of a convolutional neural network is multi-layered, with complex workflows that are difficult to navigate. In this paper, the workflow of YOLOv2 object recognition algorithm is visualizated. The D3.js JavaScript Library is used to support SVG format.We developed a system that allows user interaction. It incorporats the YOLOv2 object recognition process using a web browser. Users can understand complex workflow of the algorithm, through the interactive web pages. In addition, when adjusting the parameters on the page, the image will be updated in real time, enhancing the learning experiences. At the same time, graphical representation can speed up the understanding of the algorithm. Experienced users can use the proposed system to observe the problems and reduce the amount of time spent on adjusting the parameters. The system allows learners to get started to understand object recognition.

摘 要 i
ABSTRACT ii
致 謝 iii
目 錄 iv
表目錄 vi
圖目錄 vii
第1章 導論 1
1.1 研究背景 1
1.2 研究動機 2
1.3 論文貢獻 2
1.4 論文架構 3
第2章 相關文獻與技術 4
2.1 卷積神經網路 4
2.2 圖像視覺化 6
第3章 視覺化系統 10
3.1 研究目標 10
3.2 研究方法 10
3.3 系統架構 10
3.4 系統操作 12
3.4.1 特徵圖產生 13
3.4.2 權重圖產生 15
3.5 系統介紹 22
3.5.1 YOLOv2網路框架視覺化 23
3.5.2 卷積運算視覺化 25
3.5.3 錨框視覺化 26
第4章 結果與討論 29
4.1 系統結果 29
4.2 限制 33
第5章 結論與未來展望 34
5.1 結論 34
5.2 未來展望 34
參考文獻 35

[1]Convolution Neural Network. [Image]. Retrieved from https://chtseng.files.wordpress.com/2017/09/4293_jd_lv-5awa.png?w=1140&h=463
[2]Salakhutdinov, R., Mnih, A., & Hinton, G. (2007, June). Restricted Boltzmann machines for collaborative filtering. In Proceedings of the 24th international conference on Machine learning (pp. 791-798).
[3]Redmon, J., Divvala, S., Girshick, R., & Farhadi, A. (2016). You only look once: Unified, real-time object detection. In Proceedings of the IEEE conference on computer vision and pattern recognition (pp. 779-788).
[4]Redmon, J., & Farhadi, A. (2017). YOLO9000: better, faster, stronger. In Proceedings of the IEEE conference on computer vision and pattern recognition (pp. 7263-7271).
[5]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).
[6]Fukushima, K., & Miyake, S. (1982). Neocognitron: A self-organizing neural network model for a mechanism of visual pattern recognition. In Competition and cooperation in neural nets (pp. 267-285). Springer, Berlin, Heidelberg.
[7]LeCun, Y., Bottou, L., Bengio, Y., & Haffner, P. (1998). Gradient-based learning applied to document recognition. Proceedings of the IEEE, 86(11), 2278-2324.
[8]Bouvrie, J. (2006). Notes on convolutional neural networks.
[9]Chellapilla, K., Puri, S., & Simard, P. (2006, October). High performance convolutional neural networks for document processing.
[10]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).
[11]Simonyan, K., & Zisserman, A. (2014). Very deep convolutional networks for large-scale image recognition. arXiv preprint arXiv:1409.1556.
[12]He, K., Zhang, X., Ren, S., & Sun, J. (2016). Deep residual learning for image recognition. In Proceedings of the IEEE conference on computer vision and pattern recognition (pp. 770-778).
[13]Erhan, D., Bengio, Y., Courville, A., & Vincent, P. (2009). Visualizing higher-layer features of a deep network. University of Montreal, 1341(3), 1.
[14]Zeiler, M. D., Taylor, G. W., & Fergus, R. (2011, November). Adaptive deconvolutional networks for mid and high level feature learning. In 2011 International Conference on Computer Vision (pp. 2018-2025). IEEE.
[15]Zeiler, M. D., & Fergus, R. (2014, September). Visualizing and understanding convolutional networks. In European conference on computer vision (pp. 818-833). Springer, Cham.
[16]Long, J., Shelhamer, E., & Darrell, T. (2015). Fully convolutional networks for semantic segmentation. In Proceedings of the IEEE conference on computer vision and pattern recognition (pp. 3431-3440).
[17]Yosinski, J., Clune, J., Nguyen, A., Fuchs, T., & Lipson, H. (2015). Understanding neural networks through deep visualization. arXiv preprint arXiv:1506.06579.
[18]Harley, A. W. (2015, December). An interactive node-link visualization of convolutional neural networks. In International Symposium on Visual Computing (pp. 867-877). Springer, Cham.
[19]anchor box. [Image]. Retrieved from https://2.bp.blogspot.com/-_R-w_tWHdzc/WzJPsol7qFI/AAAAAAABbgg/Jsf-AO3qH0A9oiCeU0LQxN-wdirlOz4WgCLcBGAs/s1600/%25E8%259E%25A2%25E5%25B9%2595%25E5%25BF%25AB%25E7%2585%25A7%2B2018-06-26%2B%25E4%25B8%258B%25E5%258D%258810.36.51.png
[20]3d barchart created with d3-3d. Retrieved 25 February 2020, from https://bl.ocks.org/Niekes/613d43d39372f99ae2dcea14f0f90617