ارائه یک الگوریتم آبشاری برای بهبود سرعت و دقت یک سیستم شناسایی نوع و مدل خودرو
محورهای موضوعی : مهندسی برق و کامپیوترمحسن بيگلري 1 * , سیدعلی سلیمانی 2 , حمید حسنپور 3
1 - دانشگاه صنعتی شاهرود
2 - دانشگاه صنعتی شاهرود
3 - دانشگاه صنعتی شاهرود
کلید واژه: شناسایی دانهریز اشیا شناسایی نوع و مدل وسیله نقلیه VMMR پردازش آبشاری الگوریتم آبشاری,
چکیده مقاله :
در دهه اخیر، مطالعات بسیاری بر روی طبقهبندی دانهریز اشیا صورت گرفته است. در این نوع طبقهبندی گروه کلی شیء مشخص بوده و هدف تعیین زیرگروه دقیق آن است و شناسایی نوع و مدل وسیله نقلیه (VMMR) نیز در این حوزه قرار میگیرد. این مسئله به دلیل وجود تعداد کلاسهای زیاد، تفاوت درون کلاسی بسیار و تفاوت بین کلاسی کم از مسایل طبقهبندی دشوار به شمار میرود. علاوه بر این معمولاً سرعت سیستمهای شناسایی اشیا با افزایش دقت، کاهش مییابد و چنان که میبینیم یکی از چالشهای مهم در شبکههای عصبی عمیق به عنوان یک ابزار قدرتمند بینایی ماشین، سرعت پردازش است. در این مقاله ابتدا روشی مبتنی بر بخش برای شناسایی نوع و مدل خودرو مختصراً معرفی میگردد و سپس یک الگوریتم آبشاری برای بهبود توأمان سرعت و دقت این سیستم ارائه میشود. الگوریتم آبشاری پیشنهادی، طبقهبندهای موجود در سیستم را به صورت ترتیبی به تصویر ورودی اعمال کرده تا از حجم پردازش بکاهد. چند معیار مناسب برای رسیدن به یک ترتیب کارا از طبقهبندها معرفی شده و در نهایت ترکیبی از آنها در الگوریتم پیشنهادی به کار گرفته شده است. نتیجه آزمایشات انجامشده بر روی مجموعه داده کاملاً متفاوت BVMMR و CompCars، نشان از دقت بالای سیستم شناسایی نوع و مدل خودرو دارد. پس از اعمال الگوریتم آبشاری به سیستم مورد بحث، سرعت پردازش تا 80% افزایش یافته است در حالی که دقت سیستم تنها کاهش جزئی داشته است.
In the last decade, many researches have been done on fine-grained recognition. The main category of the object is known in this problem and the goal is to determine the subcategory or fine-grained category. Vehicle Make and Model Recognition (VMMR) is a hard fine-grained classification problem, due to the large number of classes, substantial inner-class and small inter-class distance. Furthermore, improving system accuracy leads to increasing in processing time. As we can see the state-of-the-art machine vision tool like convolutional neural networks lacks in real-time processing time. In this paper, a method has been presented briefly for VMMR firstly. Secondly, a cascading scheme for improving both speed and accuracy of this VMMR system has been proposed. In order to eliminate extra processing cost, the proposed cascading scheme applies classifiers to the input image in a sequential manner. Some effective criterions for an efficient ordering of classifiers are proposed and finally a fusion of them is used in the cascade algorithm. For evaluation purposes, a new dataset with more than 5000 vehicles of 28 different makes and models has been collected. The experimental results on this dataset and comprehensive CompCars dataset show outstanding performance of our approach. Our cascading scheme results up to 80% increase in the system processing speed.
[1] A. Krizhevsky, I. Sutskever, and G. E. Hinton, "Imagenet classification with deep convolutional neural networks," in Advances in Neural Information Processing Systems, pp. 1097-1105, 2012.
[2] P. Sermanet, D. Eigen, X. Zhang, M. Mathieu, R. Fergus, and Y. LeCun, OverFeat: Integrated Recognition, Localization and Detection Using Convolutional Networks, arXiv preprint arXiv: 1312.6229, 2013.
[3] Y. Zhou, L. Liu, L. Shao, and M. Mellor, DAVE: A Unified Framework for Fast Vehicle Detection and Annotation, arXiv: 1607.04564v2, 16 pp. , 2016.
[4] 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, no. 6, pp. 1137-1149, Jun. 2017.
[5] Y. Jia, E. Shelhamer, J. Donahue, S. Karayev, J. Long, R. Girshick, S. Guadarrama, and T. Darrell, "Caffe: convolutional architecture for fast feature embedding," in Proc. of the 22nd ACM Int.Conf. on Multimedia, pp. 675-678, Nov. 2014.
[6] S. Tokui, K. Oono, S. Hido, and J. Clayton, "Chainer: a next-generation open source framework for deep learning," in Proc. of Workshop on Machine Learning Systems (LearningSys) in the 29th Annual Conf. on Neural Information Processing Systems, 6 pp., 2015.
[7] A. Vedaldi and K. Lenc, "Matconvnet: convolutional neural networks for matlab," in Proc. of the 23rd ACM Int. Conf. on Multimedia, pp. 689-692, Oct. 2015.
[8] M. Abadi, et al., "Tensorflow: a system for large-scale machine learning," in Proc. s of the 12th USENIX Conf. on Operating Systems Design and Implementation, OSDI'16, pp. 265-283, 2-4 Nov. 2016.
[9] S. Tulyakov, S. Jaeger, V. Govindaraju, and D. Doermann, "Review of classifier combination methods," Studies in Computational Intelligence, vol. 90, pp. 361-386, 2008.
[10] P. F. Felzenszwalb, R. B. Girshick, and D. McAllester, "Cascade object detection with deformable part models," in Proc. IEEE Computer Society Conf. on Computer Vision and Pattern Recognition, pp. 2241-2248, Jun. 2010.
[11] P. Viola and M. Jones, "Rapid object detection using a boosted cascade of simple features," in Proc. IEEE Computer Society Conf. on Computer Vision and Pattern Recognition, vol. 1, pp. 511-518, Dec. 2001.
[12] B. Zhang, "Reliable classification of vehicle types based on cascade classifier ensembles," IEEE Trans. on Intelligent Transportation Systems, vol. 14, no. 1, pp. 322-332, Mar. 2013.
[13] M. Biglari, A. Soleimani, and H. Hassanpour, "Part-based recognition of vehicle make and model," IET Image Processing, vol. 11, no. 7, pp. 483-491, Mar. 2017.
[14] L. Yang, P. Luo, C. C. Loy, and X. Tang, "A large-scale car dataset for fine-grained categorization and verification," in Proc. IEEE Conf. on Computer Vision and Pattern Recognition, pp. 3973-3981, Jun. 2015.
[15] Multimedia Processing Lab, NTOU-MMR Dataset, http://mmplab.cs.ntou.edu.tw/mmplab/MMR/MMR.html (Accessed: Oct. 2016).
[16] J. Krause, H. Jin, J. Yang, and L. Fei-Fei, "Fine-grained recognition without part annotations," in Proc. IEEE Conf. on Computer Vision and Pattern Recognition, pp. 5546-5555, Jun. 2015.
[17] P. F. Felzenszwalb, R. B. Girshick, D. McAllester, and D. Ramanan, "Object detection with discriminatively trained part-based models," IEEE Trans. on Pattern Analysis and Machine Intelligence, vol. 32, no. 9, pp. 1627-1645, Sept. 2010.
[18] P. A. Savalle, S. Tsogkas, G. Papandreou, and I. Kokkinos, "Deformable part models with CNN features," in Proc. European Conf. on Computer Vision, Parts and Attributes Workshop, 5 pp., Sept. 2014.
[19] R. Girshick, F. Iandola, T. Darrell, and J. Malik, "Deformable part models are convolutional neural networks," in Proc. Proc. IEEE Conf. on Computer Vision and Pattern Recognition, pp. 437-446, Jun. 2015.
[20] –, The PASCAL Visual Object Classes, 2008, http://pascallin.ecs.soton.ac.uk/challenges/VOC/ (Accessed: 10 Mar. 2015).
[21] P. Felzenszwalb, D. McAllester, and D. Ramanan, "A discriminatively trained, multiscale, deformable part model," in IEEE Conf. on Computer Vision and Pattern Recognition, 8 pp., Jun. 2008.
[22] J. Platt, "Probabilistic outputs for support vector machines and comparisons to regularized likelihood methods," Advances in Large Margin Classifiers, vol. 10, no. 3, pp. 61-74, Mar. 1999.
[23] J. W. Hsieh, L. C. Chen, and D. Y. Chen, "Symmetrical SURF and its applications to vehicle detection and vehicle make and model recognition," IEEE Trans. on Intelligent Transportation Systems, vol. 15, no. 1, pp. 6-20, Feb. 2014.
[24] J. Fang, Y. Zhou, Y. Yu, and D. Sidan, "Fine-grained vehicle model recognition using a coarse-to-fine convolutional neural network architecture," IEEE Trans. on Intelligent Transportation Systems, vol. 18, no. 7, pp. 1782-1792, Jul. 2017.