Systems and Means of Informatics
2023, Volume 33, Issue 2, pp 13-24
RECOGNITION OF ANOMALIES ON MULTITIME PANORAMAS USING THE NEURAL NETWORK METHOD OF MODEL AMALGAMATION
- P. O. Arkhipov
- S. L. Philippskih
Abstract
A method for classifying anomalies on multitime panoramas using the neural network method of model amalgamation is described. The main idea of this method is to break the deep learning model into its component parts.
Then, existing design patterns are selected and adapted for each part. To solve the problem of classifying anomalies, a composite template was developed and a new neural network model was designed. An improvement of the design pattern of neural networks ConvNet based on the technology of using "dense convolutional blocks" has been made. To classify anomalies, a new template has been developed - SOTA-ConvNet which includes a ready-made neural network model architecture suitable for working with panoramic images. The neural network model built on the basis of the SOTA-ConvNet template showed advantages over the neural network based on ConvNet: the accuracy of image classification increased and the number of trainable parameters decreased by an order of magnitude.
[+] References (17)
- Zhang, M., J. Xie, Y. Li, and D. Wu. 2001. Color histogram correction for panoramic images. 7th Conference (International) on Virtual Systems and Multimedia Proceedings. IEEE. 328-331. doi: 10.1109/VSMM.2001.969687.
- Mohammed, A. B. E. A., F. Ming, andF. Zhengwei. 2015. Color balance for panoramic images. Modern Applied Science 9(13): 140-147. doi: 10.5539/mas.v9n13p140.
- Pakhirka, A.I., A. G. Zotin, and V.V. Buryachenko. 2018. Sozdanie panoramnykh aerofotosnimkov s ispol'zovaniem kvadrokoptera [Creating panoramic aerial images from quadcopter]. Programmnye produkty i sistemy [Software Systems] 31(2):362-367. doi: 10.15827/0236-235X.031.2.362-367.
- Arkhipov, P.O., and M.V. Tsukanov. 2020. Informatsionnaya model' tekhnologii korrektsii yarkosti i tsveta pri sozdanii panoramnykh izobrazheniy [Information model of brightness and color correction technology for creating panoramic images]. Sis- temy vysokoy dostupnosti [Highly Available Systems] 16(3):46-51. doi: 10.18127/ j20729472-202003-04.
- Arkhipov, P. O., and M. V. Tsukanov. 2021. Algoritmicheskaya model' obnaruzheniya anomaliy na raznovremennykh panoramakh [Algorithmic model of anomaly detection on different panoramas]. Sistemy vysokoy dostupnosti [Highly Available Systems] 17(2):5-10. doi: 10.18127/j20729472-202102-01.
- Arkhipov, P.O., A. K. Trofimenkov, M.V. Tsukanov, and N. Yu. Nosova. 2022. Issledovanie metodov detektirovaniya klyuchevykh tochek pri sozdanii panoramnykh izobrazheniy [Study of methods for detecting key points when creating panoramic images]. Sistemy i Sredstva Informatiki - Systems and Means of Informatics 32(2):92-104. doi: 10.14357/08696527220209.
- Arkhipov, P. O., and S. L. Philippskih. 2022. Building an ensemble of convolutional neural networks for classifying panoramic images. Pattern Recognition Image Analysis 32(3):511-514. doi: 10.1134/S 1054661822030051.
- Krizhevsky, À., I. Sutskever, and G. Hinton. 2017. ImageNet classification with deep convolutional neural networks. Commun. ACM 60:84-90. doi:10.1145/3065386.
- Ferlitsch, A. 2021. Deep learning patterns and practices. Manning Publications Co. 472 p.
- Gao, H., Zh. Liu, L. van der Maaten, and K. Q. Weinberger. 2016. Densely connected convolutional networks. arXiv.org. 9 p. Available at: https://arxiv.org/abs/ 1608.06993 (accessed April 8, 2023).
- Simonyan, K., and A. Zisserman. 2014. Very deep convolutional networks for large- scale image recognition. arXiv.org. 14p. Availableat: https://arxiv.org/abs/1409.1556 (accessed April 8, 2023).
- Howard, A., M. Zhu, Bo Chen, et al. 2017. MobileNets: Efficient convolutional neural networks for mobile vision applications. arXiv.org. 9 p. Available at: https:// arxiv.org/abs/1704.04861 (accessed April 8, 2023).
- Chollet, F. 2021. Deep learning with Python. 2nd ed. Shelter Island, NY: Manning Publications Co. 504 p.
- Philippskih, S. L. 2022. Klassifikatsiya izvlekaemykh iz panoram izobrazheniy neyronnoy set'yu s modulem sdavlivaniya-vozbuzhdeniya [Classification of images extracted from panoramas using a neural network with a squeeze-excitation module]. Intellectualization of Information Processing: 14th Conference (International) Abstracts. Moscow: Russian Academy of Sciences. 204-209.
- He, K., X. Zhang, Sh. Ren, and J. Sun. 2015. Deep residual learning for image recognition. arXiv.org. 12 p. Available at: https://arxiv.org/abs/1512.03385 (accessed April 8, 2023).
- Zhu, P., L. Wen, D. Du, et al. 2020. Detection and tracking meet drones challenge. arXiv.org. 20 p. Available at: https://arxiv.org/abs/2001.06303 (accessed April 8, 2023).
- Kingma, D., and J. B. Adam. 2014. A method for stochastic optimization. arXiv.org. 15 p. Available at: https://arxiv.org/abs/1412.6980 (accessed April 8, 2023).
[+] About this article
Title
RECOGNITION OF ANOMALIES ON MULTITIME PANORAMAS USING THE NEURAL NETWORK METHOD OF MODEL AMALGAMATION
Journal
Systems and Means of Informatics
Volume 33, Issue 2, pp 13-24
Cover Date
2023-06-10
DOI
10.14357/08696527230202
Print ISSN
0869-6527
Publisher
Institute of Informatics Problems, Russian Academy of Sciences
Additional Links
Key words
panoramic image; model amalgamation; multiclass classification; design pattern; convolutional neural network; data set
Authors
P. O. Arkhipov  and S. L. Philippskih
Author Affiliations
Federal Research Center "Computer Science and Control", Russian Academy of Sciences, 44-2 Vavilov Str., Moscow 119333, Russian Federation
|