印刷電路板是當今電子產品的基礎材料，也是電子相關產品不可或缺的重要組件。電路板完成後及插件後都需要個別檢測瑕疵，本研究只考慮插件後的元件檢測項目。在過去的傳統檢測方法中，需要針對特定元件設計出特定合適的演算法，因此檢測速度非常快，但缺點是泛用性很低，對於每種新的元件都得設計新的演算法。近年來深度學習發展快速，且廣泛的應用在各種領域，也漸漸取代了傳統方法。在本研究中，我們使用深度學習的方式來定位及分類印刷電路板上的元件。另外，有些印刷電路板上的元件是有不同方向，為了達到更準確的定位效果，我們發展了旋轉偵測框的定位功能。
本研究分為三部份，第一部份是針對標記的修改，印刷電路板上的元件允許有360度的旋轉，我們重新設計訓練影像上的標記；第二部份是針對網路的修改，我們修改YOLOv3的網路架構，以符合新的標記，我們也加入自己設計的角度相關之損失函數，並將GIoU (generalized intersection over union) 及焦點損失函數 (focal loss) 加入原本的損失函數中以加強偵測的效果；第三部份我們使用了資料前處理、標記平滑化、學習率策略及隨機輸入影像尺度的訓練策略，使新的網路能夠對印刷電路板上的元件檢測及估計更加準確。
在實驗中分為有角度估計及無角度估計的偵測網路，使用的訓練資料為印刷電路板元件資料集，共有298張影像。無角度估計的偵測網路中，其中267張影像用於訓練，31張影像用於測試；有角度估計的偵測網路中，非0度元件資料過少，298張影像皆用於訓練，無測試影像。偵測與辨識的元件共分26類。
在實驗中，無角度估計的偵測網路僅加入兩種損失，且應用四種訓練策略，其中訓練資料的mAP達到99.9%，測試資料的mAP達到89.3%。有角度估計的偵測網路中，亦將訓練資料加入角度的資訊，且在損失函數中加入角度損失，其中訓練資料的mAP達到90.4%。

The printed circuit board is the basic and irreplaceable component of electronic products nowadays. There is generally a component detection at the later stage of the component-assembly process for detecting defect. In the past, it was necessary to design a specific suitable algorithm for specific components. This algorithm will be very fast, but the versatility is very low. It means that a new algorithm should be designed for each new component everytimes. Deep learning has developed rapidly, and is widely used in various fields in recent. It has gradually replaced traditional manually algorithms. In this study, we used deep learning to classify and estimate position of components on the printed circuit board. On the printed circuit board, some components have different positions. In order to get more precise position, we have developed a method for these components.
There are three parts in our research. First, We redesigned the label of training image for the components on the printed circuit board which have angle. Second, we modified the network structure of YOLOv3 to fit the new label. We also used angle-related, generalized intersection over union and focal loss function to strengthen the effect of detection. Finally, we used some training strategies witch are label smoothing, data pre-processing, learning rate schedule and random shapes training let the network structure get more accuracy for the components on the printed circuit board.
In experiment, we used detection networks whether have angle estimation or not. The printed circuit board component dataset be used as training data which have 298 images. In non-angle estimation detection network, there are 267 images used in training and 31 images used in testing. In angle estimation detection network, there are 298 images used in training and no image used in testing. There are 26 classes in our dataset.
We modified two loss functions and used four training strategies in non-angle estimation detection networks. The mAP of training data reaches 99.9%, and the mAP of testing data reaches 89.3%. We redesigned the label of images and used angle-related loss function in angle estimation detection networks. The mAP of training data finally reached 90.4%.

[1] A. Krizhevsky, I. Sutskever, and G. Hinton, “Imagenet classification with deep convolutional neural networks,” in Proc. of Neural Information Processing Systems (NIPS), Harrahs and Harveys, Lake Tahoe, NV, Dec.3-8, 2012.
[2] P. Sermanet, D. Eigen, X. Zhang, M. Mathieu, R. Fergus, and Y. Le-Cun, “Overfeat: Integrated recognition, localization and detection using convolutional networks,” in Proc. of ICLR Conf., Banff, Canada, Apr.14-16, 2014.
[3] M. D. Zeiler and R. Fergus, “Visualizing and understanding convolutional neural networks,” in Proc. of ECCV Conf., Zurich, Switzerland, Sep.6-12, 2014, pp.818-833.
[4] K. Simonyan and A. Zisserman, “Very deep convolutional networks for large-scale image recognition,” arXiv:1409.1556.
[5] C. Szegedy, W. Liu, Y. Jia, P. Sermanet, S. Reed, D. Anguelov, D. Erhan, V. Vanhoucke, and A. Rabinovich, “Going deeper with convolutions,” in Proc. of IEEE Conf. on Computer Vision and Pattern Recognition (CVPR), Boston, MA, Jun.7-12, 2015, pp.1-9.
[6] J. Redmon and A. Farhadi, “YOLOv3: an incremental improvement,” arXiv:1804.02767.
[7] S. Ioffe and C. Szegedy, “Batch normalization: accelerating deep network training by reducing internal covariate shift,” in Proc. of ICML Conf. , Lille, France, Jul.7-9, 2015, vol.37, pp.448-456.
[8] K. He, X. Zhang, S. Ren, and J. Sun, “Deep residual learning for image recognition,” in Proc. of IEEE Conf. on Computer Vision and Pattern Recognition (CVPR), Las Vegas, NV, Jun.27-30, 2016, pp.770-778.
[9] H. Rezatofighi, N. Tsoi, J. Gwak, A. Sadeghian, I. Reid, S. Savarese, “Generalized intersection over union: a metric and a loss for bounding box regression,” in Proc. of IEEE Conf. on Computer Vision and Pattern Recognition (CVPR), CA, June.16-20, 2019, pp.658-666.
[10] A. Neubeck, L. V. Gool, “Efficient non-maximum suppression,” in Proc. of 18th Int. Conf. on Pattern Recognition (ICPR), Hong Kong, Aug.20-24, 2006, pp.850-855.
[11] R. Girshick, J. Donahue, T. Darrell, and J. Malik, “Rich feature hierarchies for accurate object detection and semantic segmentation,” in Proc. IEEE Conf. on Computer Vision and Pattern Recognition (CVPR), Columbus, Ohio, Jun.23-28, 2014, pp.580-587.
[12] J. Uijlings, K. Sande, T. Gevers, and A. Smeulders, “Selective search for object recognition,” Int. Journal of Computer Vision (IJCV), vol.104, is.2, pp.154-171, 2013.
[13] R. Girshick, “Fast R-CNN,” in Proc. of IEEE Int. Conf. on Computer Vision (ICCV), Santiago, Chile, Dec.11-18, 2015, pp.1440-1448.
[14] K. He, X. Zhang, S. Ren, and J. Sun, “Spatial pyramid pooling in deep convolutional networks for visual recognition,” in Proc. of ECCV Conf. , Zurich, Switzerland, Sep.6-12, 2014, pp.346-361.
[15] S. Ren, K. He, R. Girshick, and J. Sun, “Faster R-CNN： Towards real-time object detection with region proposal networks,” IEEE Trans. on Pattern Analysis and Machine Intelligence, vol.39, is.6, pp.1137-1149, 2016.
[16] K. He, G. Gkioxari, P. Dollár and R. Girshick, “Mask R-CNN,” in Proc. of IEEE Int. Conf. on Computer Vision (ICCV), Venice, Italy, Oct.22-29, 2017, pp. 2980-2988.
[17] J. Redmon, S. Divvala, R. Girshick and A. Farhadi, “You only look once: unified, real-time object detection,” in Proc. of IEEE Conf. on Computer Vision and Pattern Recognition (CVPR), Las Vegas, NV, Jun.27-30, 2016, pp.779-788.
[18] W. Liu, D. Anguelov, D. Erhan, C. Szegedy, S. Reed, C.-Y. Fu, and A. C. Berg, “ SSD: Single shot multibox detector,” in European Conf. on Computer Vision (ECCV), Amsterdam, Holland, Oct.8-16, 2016, pp.21-37.
[19] C.-Y. Fu, W. Liu, A. Ranga, A. Tyagi, and A. C. Berg, “Dssd: Deconvolutional single shot detector,” arXiv:1701.06659.
[20] J. Redmon and A. Farhadi, “YOLO9000: better, faster, stronger,” in Proc. of IEEE Conf. on Computer Vision and Pattern Recognition (CVPR), Honolulu, HI, Jul.21-26, 2017, pp.6517-6525.
[21] T.-Y. Lin, P. Dollár1, R. Girshick, K. He, B. Hariharan, and S. Belongie, “Feature pyramid networks for object detection,” in Proc. of IEEE Conf. on Computer Vision and Pattern Recognition (CVPR), Honolulu, HI, Jul.21-26, 2017, pp.936-944.
[22] C. Szegedy, V. Vanhoucke, S. Ioffe, J. Shlens, and Z. Wojna, “Rethinking the inception architecture for computer vision,” in Proc. of IEEE Conf. on Computer Vision and Pattern Recognition (CVPR), Las Vegas, NV, Jun.27-30, 2016, pp.2818-2826.
[23] P. Goyal, P. Dollár, R. Girshick, P. Noordhuis, L. Wesolowski, A. Kyrola, A. Tulloch, Y. Jia, and K. He, “Accurate, large minibatch sgd: training imagenet in 1 hour,” arXiv:1706.02677.
[24] Z. Zhang, T. He, H. Zhang, Z. Zhang, J. Xie, and M. Li, “Bag of freebies for training object detection neural networks,” arXiv:1902.04103.
[25] I. Loshchilov and F. Hutter, “SGDR: stochastic gradient descent with warm restarts,” arXiv:1608.03983.
[26] H. Zhang, M. Cisse, Y. N. Dauphin, and D. L.-Paz. “Mixup: beyond empirical risk minimization,” arXiv:1710.09412.
[27] Z. Zhang, T. He, H. Zhang, Z. Zhang, J. Xie, and M. Li, “Bag of tricks for image classification with convolutional neural networks,” arXiv:1812.01187.
[28] A. Rosenfeld, R. Zemel, and J. K. Tsotsos, “The elephant in the room, ” arXiv:1808.03305.
[29] C. Peng, T. Xiao, Z. Li, Y. Jiang, X. Zhang, K. Jia, G. Yu, and J. Sun, “Megdet: A large mini-batch object detector,” in Proc. of IEEE Conf. on Computer Vision and Pattern Recognition (CVPR), Salt Lake City, UT, Jun.18-23, 2018, pp.6181-6189.
[30] H. Zhang, K. Dana, J. Shi, Z. Zhang, X.Wang, A. Tyagi, and A. Agrawal, “Context encoding for semantic segmentation,” in Proc. of IEEE Conf. on Computer Vision and Pattern Recognition (CVPR), Salt Lake City, UT, Jun.18-23, 2018, pp.7151-7160.
[31] Z. Zheng, P. Wang, W. Liu, J. Li, R. Ye, D. Ren, “Distance-IoU loss: faster and better learning for bounding box regression,” arXiv:1911.08287.
[32] R. K. Srivastava, K. Greff, and J. Schmidhuber, “Training very deep networks,” arXiv:1507.06228.
[33] M. Simon, S. Milz, K. Amende, and H.-M. Gross, “Complex-YOLO: real-time 3D object detection on point clouds,” arXiv:1803.06199.
[34] M. Simon, K. Amende, A. Kraus, J. Honer, T. Sämann, H. Kaulbersch, S. Milz, and H. M. Gross, “Complexer-YOLO: Real-time 3D object detection and tracking on semantic point clouds,” arXiv:1904.07537.
[35] M. A. Rahman and Y. Wang, eds.“Optimizing intersection-overunion in deep neural networks for image segmentation,” in Int. Symp. on Visual Computing, Las Vegas, NV, Dec.12-14, 2016, pp.234-244.
[36] T.-Y. Lin, P. Goyal, R. Girshick, K. He, and P. Dollár, “Focal loss for dense object detection,” arXiv:1708.02002.