基於邊緣運算進行深度學習物件偵測之效能評估｜國立臺灣科技大學博碩士論文系統

簡易檢索 / 詳目顯示

回結果列表

研究生：	陳譔文 Chuan-Wen Chen
論文名稱：	基於邊緣運算進行深度學習物件偵測之效能評估 Performance Evaluation of Edge Computing-Based Deep Learning Object Detection
指導教授：	阮聖彰 Shanq-Jang Ruan 林昌鴻 Chang-Hong Lin
口試委員:	吳晉賢 Chin-Hsien Wu 林淵翔 Yuan-Hsiang Lin
學位類別：	碩士 Master
系所名稱：	電資學院 - 電子工程系 Department of Electronic and Computer Engineering
論文出版年：	2018
畢業學年度：	107
語文別：	中文
論文頁數：	65
中文關鍵詞：	深度學習、物件偵測、物聯網裝置、邊緣運算
外文關鍵詞：	Deep Learning, Object Detection, IoT Device, Edge Computing
相關次數：	點閱：352 下載：5
分享至:	分享至facebook 分享至twitter

查詢本校圖書館目錄查詢臺灣博碩士論文知識加值系統勘誤回報

上一筆

本篇研究論文提出了一種基於邊緣運算進行深度學習之物件偵測實現在低成本的物聯網裝置之方法。由於低成本物聯網裝置硬體上運算的限制，要運行訓練好的神經網絡模型將是一項挑戰，因此本篇研究論文利用Intel旗下公司Movidius所開發的神經運算棒(Neural Compute Stick, NCS)，藉由神經運算棒高效的浮點數運算能力，將其結合於低成本之物聯網裝置上，使低成本的物聯網裝置能順利運行已經訓練好的神經網絡模型，成為邊緣運算物聯網裝置。而經由實驗結果顯示出本篇研究論文所提出之方法能有效地在邊緣端物聯網裝置上運行深度學習之物件偵測，驗證出神經運算棒在邊緣端物聯網裝置對於靜態影像推理運行神經網絡模型的時間能夠在1.7秒內完成，而對於動態影片加速推理運行時間也能夠達到平均9.2 FPS。

This thesis presents a method for implementing deep learning object detection based on low-cost edge computing IoT (Internet-of-Things) device. The limit of hardware is a challenge for working the pre-trained neural network model on the low-cost IoT device. Therefore, this thesis utilizes the Intel Movidius Neural Compute Stick (NCS) to accel-erate the neural network model on the low-cost IoT device by its high efficient floating point operation. With the Neural Compute Stick, the low-cost IoT device can success-fully work the pre-trained neural network model and become an edge computing device. The experimental results show the proposed method can effectively detect the objects based on deep learning on the edge-side IoT device. Furthermore, the objective experi-ment demonstrates the proposed method can immediately infer the neural network model in average 1.7 seconds。

目錄
第一章 緒論    1
1-1前言    1
1-2動機與論文目標    2
1-3論文架構    3
第二章 背景文獻    4
2-1卷積神經網絡    4
2-2深度學習物件偵測方法    7
2-2-1 R-CNN(Regions with CNN)    8
2-2-2 Fast R-CNN(Fast Regions with CNN)    9
2-2-3 YOLO(You Only Look Once)    11
2-3本論文實現之方法    13
2-3-1 SSD(Single Shot MultiBox Detector)    13
2-3-2 MobileNets    17
2-4相關深度學習邊緣運算文獻    19
第三章 研究方法探討    21
3-1 收集資料數據集    23
3-1-2 KITTI    24
3-1-3 自製數據集LPS2017    24
3-2 網絡模型訓練    25
3-3 網絡模型應用於邊緣裝置    27
3-4 網絡模型更新    29
第四章 實驗結果    31
4-1 硬體環境    31
4-2 物件偵測準確度評估    32
4-2-1 交疊區域(Intersection over Union, IoU)    32
4-2-2 平均精確度(Mean Average Precision, mAP)    33
4-3 物件偵測範例偵測    39
4-4 物件偵測推理運行時間評估    43
4-4-1 靜態影像推理運行時間評估    44
4-4-2 動態影片推理運行時間評估    45
4-5 網絡模型更新精確度評估    48
第五章 結論與未來展望    49
參考文獻    50


圖目錄
圖1-1：邊緣運算示意圖    2
圖2-1：卷積神經網絡架構    4
圖2-2：卷積運算步驟    5
圖2-3：最大池化運算步驟    6
圖2-4：(a) 物件辨識。(b) 物件偵測。    7
圖2-5：R-CNN物件偵測方法    8
圖2-6：Fast R-CNN物件偵測方法    9
圖2-7：ROI pooling方法。(a) 輸入特徵圖。(b) 黑線預選框映射特徵圖上。(c) 各區域最大池化值。(d) 輸出結果。    10
圖2-8：YOLO物件偵測方法    12
圖2-9：YOLO網絡架構    12
圖2-10：SSD模型和YOLO模型之比較[13]    13
圖2-11：利用3×3卷積遮罩提取特徵    14
圖2-12：(a) Ground truth原圖。(b) 8×8特徵圖。(c) 4×4特徵圖。    15
圖2-13：標準卷積遮罩模型    17
圖2-14：深度卷積遮罩    18
圖2-15：點卷積遮罩    18
圖3-1：本論文所提出之研究方法流程    22
圖3-2：PASCAL VOC數據集    23
圖3-3：KITTI數據集    24
圖3-4：自製數據集LPS2017    25
圖3-5：(a) VGG-16標準卷積。(b) 深度卷積和1×1大小點卷積。    25
圖3-6：BBox-Label-Tool工具    26
圖3-7：樹莓派Raspberry Pi 3    27
圖3-8：Movidius神經運算棒(Neural Compute Stick, NCS)    28
圖3-9：邊緣端物聯網裝置架設網路攝影機    29
圖3-10：輸出資料提取    30
圖4-1：交疊區域IoU    32
圖4-2：不同的IoU閥值比較    33
圖4-3：偵測精確度比較 (a)邊緣端伺服器運行之結果。(b)邊緣端物聯網裝置運行之結果。    39
圖4-4：偵測精確度比較 (a)邊緣端伺服器運行之結果。(b)邊緣端物聯網裝置運行之結果。    39
圖4-5：偵測精確度比較 (a)邊緣端伺服器運行之結果。(b)邊緣端物聯網裝置運行之結果。    40
圖4-6：偵測精確度比較 (a)邊緣端伺服器運行之結果。(b)邊緣端物聯網裝置運行之結果。    40
圖4-7：偵測精確度比較 (a)邊緣端伺服器運行之結果。(b)邊緣端物聯網裝置運行之結果。    41
圖4-8：偵測精確度比較 (a)邊緣端伺服器運行之結果。(b)邊緣端物聯網裝置運行之結果。    41
圖4-9：偵測精確度比較 (a)邊緣端伺服器運行之結果。(b)邊緣端物聯網裝置運行之結果。    42
圖4-10：偵測精確度比較 (a)邊緣端伺服器運行之結果。(b)邊緣端物聯網裝置運行之結果。    42
圖4-11：動態影片加速運算方法    43
圖4-12：靜態影像推理運行時間比較    44
圖4-13：LPS2017各大公共場所監測影片    46
圖4-14：動態影片推理運行時間比較    46
圖4-15：動態影片加速推理運行時間比較    47
圖4-16：模型更新精確度評估    48


                                

參考文獻
[1]S. Patidar and D. Rane, “A Survey Paper on Cloud Computing,” in 2012 Second Interna tional Conference on Advanced Computing & Communication Technolo-gies,pp. 394-398, Jan. 2012.
[2]S. Yi, Z. Hao and Q. Li, “Fog Computing: Platform and Applications,” in Proc. of IEEE Workshop on Hot Topics in Web Systems and Technologies (HotWeb),pp 73-78, Nov. 2015.
[3]R. T. Azuma, “A Survey of Augmented Reality,” in Presence: Teleoperators and Virtual Environments, vol. 6, no. 4, pp. 355–385, 1997.
[4]R. Girshick, J. Donahue, T. Darrell, and J. Malik, “Rich feature hierarchies for ac curate object detection and semantic segmentation,” in Proc. Conference on Computer Vision and Pattern Recognition (CVPR), pp. 580-587, Jan. 2014.
[5]R. Girshick, “Fast R-CNN,” in IEEE International Conference on Computer Vi-sion (ICCV), pp. 1440-1448, Dec. 2015.
[6]P. Sermanet, D. Eigen, X. Zhang, M. Mathieu, R. Fergus, and Y. LeCun, “Over-Feat: Integrated Recognition, Localization and Detection using Convolutional Networks,” in Proc. International Conference on Learning Representations (ICLR), Feb. 2014.
[7]A. Krizhevsky, I. Sutskever, and G. Hinton, “ImageNet Classification with Deep Convolutional Neural Networks,” in Advances in neural information processing systems (NIPS), pp. 1097–1105, 2012.
[8]D. Hubel and T. Wiesel, “Receptive fields of single neurones in the cat’s striate cortex,” J. Physiol. (London) 148, pp. 574–591, Apr. 1959.
[9]K. Fukushima, “Neocognitron: A Self-Organizing Neural Network Model for a Mechanism of Pattern Recognition Unaffected by Shift in Position,” Biological Cybernetics, vol. 36, pp. 193-202, Apr. 1980.
[10]Y. Lecun, L. Bottou, Y. Bengio and P. Haffner, “Gradient-based learning applied to document recognition,” in Proceedings of the IEEE, vol. 86, no.11, pp. 2278–2324, Nov. 1998.
[11]K. He, X. Zhang, S. Ren and J. Sun, “Deep Residual Learning for Image Recog-nition,” in IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pp. 770-778, Jun. 2016.
[12]D. Erhan, C. Szegedy, A. Toshev, and D. Anguelov, “Scalable object detection using deep neural networks,” in IEEE Conference on Computer Vision and Pat-tern Recognition (CVPR),pp. 23-28, Jun. 2014.
[13]J. Uijlings, K. van de Sande, T. Gevers and A. Smeulders, “Selective Search for Object Recognition,” in International Journal of Computer Vision (IJCV), vol. 104, pp. 154-171, Sept. 2013.
[14]K. He, X. Zhang, S. Ren and J. Sun, “Spatial Pyramid Pooling in Deep Convolu-tional Networks for Visual Recognition,” in IEEE Transactions on Pattern Analy-sis and Machine Intelligence, vol. 37, no. 9, pp. 1904-1916, Sept. 2015.
[15]J. Redmon, S. Divvala, R. Girshick, A. Farhadi, “You only look once: unified, re-al-time object detection,” in IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pp. 779-788, Jun. 2016.
[16]W. Liu, D. Anguelov, D. Erhan, C. Szegedy, S. Reed, C. Y. Fu and A. C. Berg, “SSD: Single Shot Multibox Detector,” in European Conference on Computer Vi-sion, pp. 21–37, Springer, 2016.
[17]A. Howard, M. Zhu, B. Chen and D. Kalenichenko, “MobileNets: Efficient Con-volutional Neural Networks for Mobile Vision Applications,” arXiv preprint arXiv :1704.04861, 2017.
[18]K. Simonyan and A. Zisserman, “Very Deep Convolutional Networks for Large-scale Image Recognition,” CoRR, abs/1409.1556, 2014.
[19]C. Liu, Yu. Cao, Y. Luo, G. Chen, V. Vokkarane, M. Yunsheng, S. Chen and P. Hou, “A New Deep Learning-Based Food Recognition System for Dietary As-sessment on an Edge Computing Service Infrastructure,” in IEEE Transactions on Services Computing, vol. 11, no. 2, pp. 249-261, Jan. 2017.
[20]M. Hosseini, T. Tran, D. Pompili, K. Elisevich and H. Soltanian-Zadeh, “Deep Learning with Edge Computing for Localization of Epileptogenicity using Multi-modal rs-fMRI and EEG Big Data,” in IEEE International Conference on Auto-nomic Computing, pp. 83-92, Jul. 2017.
[21]J. Ren, Y. Guo, D. Zhang, Q. Liu and Y. Zhang, “Distributed and Efficient Object Detection in Edge Computing: Challenges and Solutions,” in IEEE Network, no. 99, pp. 1-7, Apr. 2018.
[22]H. Li, K. Ota and M. Dong, “Learning IoT in edge: Deep Learning for the Internet of Things with Edge Computing,” in IEEE Network, vol. 32, no. 1,pp. 96–101, Jan. 2018.
[23]Y. Huang, X. Ma, X. Fan, J. Liu, and W. Gong, ‘‘When Deep Learning Meets Edge Computing,’’ in IEEE International Conference on Networks Protocols (ICNP), pp.1-2, Oct. 2017.
[24]C. Szegedy, W. Liu, Y. Jia, P. Sermanet, S. Reed, D. Anguelov, D. Erhan, V. Vanhoucke, A. Rabinovich, ‘‘Going deeper with convolutions,’’ in 2015 IEEE Conference on Computer Vision and Pattern Recognition(CVPR), pp. 1-9, Jun. 2015.
[25]M. Verhelst, and B, Moons, ‘‘Embedded Deep Neural Network Processing: Algo-rithmic and Processor Techniques Bring Deep Learning to IoT and Edge Devices,’’ in IEEE Solid-State Circuits Magazine, vol.9, no. 4, pp. 55-65, Nov. 2017.
[26]M. Everingham, S. M. A. Eslami, L. Van Gool, C. K. I. Williams, J. Winn and A. Zisserman, “The PASCAL Visual Object Classes Challenge: A Retrospective,” in International Journal of Computer Vision, vol. 111, no.1, pp. 98-136, Jan. 2015.
[27]G. Andreas, L. Philip, S. Christoph and U. Raquel, “Vision meets robotics: The KITTI dataset,” in International Journal of Robotics Research, pp. 1231-1237, Sept. 2013.
[28]Y. Jia, E. Shelhamer, J. Donahue, S. Karayev, J. Long, R. Girshick, S. Guadarra-ma, and T. Darrell, “Caffe: Convolutional architecture for fast feature embedding,”. in Proceedings of the 22nd ACM International Conference on Multimedia, pp. 675-678, Nov. 2014.

簡易檢索 / 詳目顯示

相關論文