Yaya Özellik Tanıma için LM Filtre Temelli Derin Evrişimsel Sinir Ağı

Günümüzde Evrişimsel Sinir Ağı (ESA) mimarileri güvenlik, endüstri ve büyük veri gibi birçok farklı alanda aktif olarak kullanılmaktadır. Bu mimarilerdeki evrişim katmanları, bir sınıflandırma veya tanımlama problemi için istenilen sonuçları verebilecek en iyi öznitelikleri otomatik olarak çıkartabilmektedir. Bu çalışmada, hem geleneksel hem de derin öznitelikleri hesaplamak için yeni bir Hibrit Evrişimsel Sinir Ağı (HESA) mimarisi önerilmiştir. Bu ağ mimarisinin temel amacı, LM filtrelerinden elde edilen geleneksel öznitelikler ile ESA mimarisinden elde edilen derin öznitelikleri birleştirerek güçlü bir öznitelik verisi oluşturmaktır. Önerilen modelde yaya görüntüsünden elde edilen LM filtre öznitelikleri ve derin öznitelikler eşzamanlı olarak hesaplanmaktadır. Daha sonra bu öznitelikler birleştirilir ve farklı öznitelikten oluşan bir öznitelik vektörü oluşturulur. Bu öznitelik vektörü tam bağlı katmanlar yardımı ile sınıflandırma işlemine alınır. Geliştirilen HESA mimarisi çok zor bir problem olan yaya özellik sınıflandırması için uygulanmıştır. Önerilen model PETA veri tabanında SVM ve MRF tabanlı yöntemlerden önemli ölçüde daha iyi performans göstermiştir. Ayrıca, ReduceLROnPlateau modelinin HESA yönteminde kullanılması yüksek başarıların elde edilmesine önemli bir katkı sağlamıştır.

Anahtar Kelimeler:

Yaya özellik tanıma, derin öğrenme, öznitelik çıkartma

LM Filter-Based Deep Convolutional Neural Network for Pedestrian Attribute Recognition

Today, Convolutional Neural Network (CNN) architectures have been used actively in many different areas such as security, industry and big data. Thanks to the convolution layers in these architectures, they can automatically extract the best features that can give the desired results for a classification or definition problem. In this paper, a new Hybrid Convolutional Neural Network (HESA) architecture is proposed to calculate both the traditional and the deep features. The main purpose of this network architecture is to combine the traditional features obtained from the LM filters and the deep features obtained from the CNN architecture so thus create a strong feature data for classification. In the proposed model, the LM filter features and deep features of the pedestrian image are calculated simultaneously. Then, these features are combined and features vector consisting of different features is built. This feature vector is taken into the classification process with the help of fully connected layer. The developed HESA architecture has been applied for the pedestrian attribute classification which is a very difficult problem. The proposed model significantly outperforms the SVM and MRF based methods on the PETA database. In addition, the use of the ReduceLROnPlateau model in the HESA method has made a significant contribution to achieving high successes.

Keywords:

Pedestrian attribute recognition, deep learning, feature extraction,

PDF

___

[1] R. Layne, T. M. Hospedales, and S. Gong, “Attributes-Based Re-identification”, Person Re-Identification, London: Springer London, 2014, pp. 93–117.
[2] M. Hirzer, C. Beleznai, P. M. Roth, and H. Bischof, “Person Re-identification by Descriptive and Discriminative Classification”, Scandinavian conference on Image analysis, Springer, Berlin, Heidelberg, 2011, pp. 91–102.
[3] L. Cao, M. Dikmen, Y. Fu, and T. S. Huang, “Gender recognition from body”, Proceeding of the 16th ACM international conference on Multimedia, New York, New York, USA: ACM Press, 2008, pp. 725–728.
[4] J. Zhu, S. Liao, Z. Lei, and S. Z. Li, “Multi-label convolutional neural network based pedestrian attribute classification”, Image Vis. Comput., vol. 58, pp. 224–229, Feb. 2017.
[5] Y. Deng, P. Luo, C. C. Loy, and X. Tang, “Pedestrian Attribute Recognition At Far Distance”, Proceedings of the ACM International Conference on Multimedia - MM ’14, New York, New York, USA: ACM Press, 2014, pp. 789–792.
[6] Y. Deng, P. Luo, C. C. Loy, and X. Tang, “Learning to Recognize Pedestrian Attribute”, arXiv Prepr. arXiv1501.00901, Jan. 2015.
[7] R. Layne, T. M. Hospedales, and S. Gong, “Towards Person Identification and Re-identification with Attributes”, European Conference on Computer Vision, Springer, Berlin, Heidelberg, 2012, pp. 402–412.
[8] T. Matsukawa and E. Suzuki, “Person re-identification using CNN features learned from combination of attributes”, 2016 23rd International Conference on Pattern Recognition (ICPR), 2016, pp. 2428–2433.
[9] E. S. Jaha and M. S. Nixon, “Soft biometrics for subject identification using clothing attributes”, IEEE International Joint Conference on Biometrics, 2014, pp. 1–6.
[10] L. An, Xiaojing Chen, M. Kafai, S. Yang, and B. Bhanu, “Improving person re-identification by soft biometrics based reranking”, 2013 Seventh International Conference on Distributed Smart Cameras (ICDSC), 2013, pp. 1–6.
[11] J. Friedman, T. Hastie, and R. Tibshirani, “Additive logistic regression: a statistical view of boosting (With discussion and a rejoinder by the authors)”, Ann. Stat., vol. 28, no. 2, pp. 337–407, Apr. 2000.
[12] J. Zhu, S. Liao, Z. Lei, D. Yi, and S. Li, “Pedestrian Attribute Classification in Surveillance: Database and Evaluation”, The IEEE International Conference on Computer Vision (ICCV) Workshops, 2013, pp. 331–338.
[13] J. Zhu, S. Liao, Z. Lei, and S. Z. Li, “Improve Pedestrian Attribute Classification by Weighted Interactions from Other Attributes”, Computer Vision - ACCV 2014 Workshops. ACCV 2014. Lecture Notes in Computer Science, 2014, pp. 545–557.
[14] L. Bourdev, S. Maji, and J. Malik, “Describing people: A poselet-based approach to attribute classification”, 2011 International Conference on Computer Vision, 2011, pp. 1543–1550.
[15] C. Su, S. Zhang, J. Xing, W. Gao, and Q. Tian, “Deep Attributes Driven Multi-Camera Person Re-identification”, European conference on computer vision, 2016, pp. 475–491.
[16] J. Zhu, S. Liao, D. Yi, Z. Lei, and S. Z. Li, “Multi-label CNN based pedestrian attribute learning for soft biometrics”, 2015 International Conference on Biometrics (ICB), 2015, pp. 535–540.
[17] P. Liu, X. Liu, J. Yan, and J. Shao, “Localization Guided Learning for Pedestrian Attribute Recognition”, Localization Guid. Learn. Pedestr. Attrib. Recognit., Aug. 2018.
[18] A. J. O’Toole, C. D. Castillo, C. J. Parde, M. Q. Hill, and R. Chellappa, “Face Space Representations in Deep Convolutional Neural Networks”, Trends Cogn. Sci., vol. 22, no. 9, pp. 794–809, Sep. 2018.
[19] Y. Seo and K. Shin, “Hierarchical convolutional neural networks for fashion image classification”, Expert Syst. Appl., vol. 116, pp. 328–339, Feb. 2019.
[20] L. A. Gatys, A. S. Ecker, and M. Bethge, “Texture and art with deep neural networks”, Curr. Opin. Neurobiol., vol. 46, pp. 178–186, Oct. 2017.
[21] K. Simonyan and A. Zisserman, “Very Deep Convolutional Networks for Large-Scale Image Recognition”, arXiv preprint arXiv:1409.1556., Sep. 2014.
[22] “CS231n Convolutional Neural Networks for Visual Recognition.” [Online]. Available: http://cs231n.github.io/convolutional-networks/#conv. [Accessed: 17-Dec-2018].