Abstract
Kee** railway tracks in correct conditions is of paramount importance to ensure adequate performance of trains, and to avoid any issues that might end up in accidents. The development of computational technology together with artificial vision techniques have boosted the use of video cameras for inspecting railway tracks. Some deep learning techniques have been applied to the day, being Convolutional Neural Networks the most popular ones. This paper presents the use of EfficientNet architecture on a railway track fault detection as a novel technique in this field. The paper compares B0 to B7 network families as well as studies the effect of input image resolution on a small dataset. On the one hand, results show that image resolution has an impact on validation accuracy and in fact, specific networks developed for particular image size are not always the best option for that image. B7 family has outperformed the rest of networks, reaching validation accuracy of 89.1%, while B2 scored 87.5% of accuracy, being the best in computational cost and performance ratio. In average, B7 and B2 have been proved to be the better solutions, with 84.6% and 84.2% accuracy respectively.
This is a preview of subscription content, log in via an institution to check access.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Similar content being viewed by others
References
International Energy Agency: The Future of Rail (2019)
Li, Q., Ren, S.: A real-time visual inspection system for discrete surface defects of rail heads. IEEE 61(8), 2189–2199 (2012)
Li, Q., Ren, S.: A visual detection system for rail surface defects. IEEE 42(6), 1531–1542 (2012)
James, A., et al.: TrackNet - a deep learning based fault detection for railway track inspection. IEEE 1, 1–5 (2018)
Yaman, O., Karaköse, M., Ak, E., Ayd, I.: Ray Yüzeyi için Görüntü İş leme Tabanl ı Ar ı za Tespit Yakla ş ı m ı Image Processing Based Fault Detection Approach for Rail Surface. In: 2015 23nd Signal Processing and Communications Applications Conference (SIU), p. 4. IEEE (2015)
Wei, X., Wei, D., Suo, D.A., Jia, L., Li, Y.: Multi-target defect identification for railway track line based on image processing and improved YOLOv3 model. IEEE Access 8, 1 (2020)
Tan, M., Le, Q.V.: EfficientNet : rethinking model scaling for convolutional neural network. In: International Conference on Machine Learning (2019)
Tumor, B., Using, C.: Brain tumor classification using dense Efficient-Net. Axioms 11(1), 13 (2022)
Ebenezer, A.S., Kanmani, S.D., Sivakumar, M., Priya, S.J.: Materials today: proceedings effect of image transformation on EfficientNet model for COVID-19 CT image classification. Mater. Today Proc. 51, 2512–2519 (2022)
Marques, G., Agarwal, D., De, I., Díez, T.: Automated medical diagnosis of COVID-19 through EfficientNet convolutional neural network. Appl. Soft Comput. J. 96, 106691 (2020)
Abdou, P., Karou, K.: Accurate detection of COVID-19 using K-EfficientNet deep learning image classifier and K-COVID chest X-ray images dataset. In: 2020 IEEE 6th International Conference on Computer and Communications (ICCC), pp. 1527–1531 (2021)
Müftüoglu, Z., Ayyüce Kizrak, M., Yildirim, T.: Differential privacy practice on diagnosis of COVID-19 radiology imaging using EfficientNet. Cornell University Library, pp. 1–6 (2020)
Kaggle, Kaggle. [Online]. https://www.kaggle.com/. Accessed 1 May 2022
López-Estrada, F.R., Méndez-López, A., Santos-Ruiz, I., Valencia-Palomo, G., Escobar-Gómez, E.: Fault detection in unmanned aerial vehicles via orientation signals and machine learning. Rev. Iberoam. Autom. Inf. Indust. 18(3), 254–264 (2021). https://doi.org/10.4995/riai.2020.14031
Márquez-Vera, M.A., López-Ortega, O., Ramos-Velasco, L.E., Ortega-Mendoza, R.M., Fernández-Neri, B.J., Zúñiga-Peña, N.S.: Fault diagnosis in industrial process by using LSTM and an elastic net. Rev. Iberoam. Autom. Inf. Indust. 18(2), 160–171 (2021). https://doi.org/10.4995/riai.2020.13611
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2022 The Author(s), under exclusive license to Springer Nature Switzerland AG
About this paper
Cite this paper
Rengel, J., Santos, M., Pandit, R. (2022). EfficientNet Architecture Family Analysis on Railway Track Defects. In: Yin, H., Camacho, D., Tino, P. (eds) Intelligent Data Engineering and Automated Learning – IDEAL 2022. IDEAL 2022. Lecture Notes in Computer Science, vol 13756. Springer, Cham. https://doi.org/10.1007/978-3-031-21753-1_46
Download citation
DOI: https://doi.org/10.1007/978-3-031-21753-1_46
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-031-21752-4
Online ISBN: 978-3-031-21753-1
eBook Packages: Computer ScienceComputer Science (R0)