Abstract
Recent advancements in cognitive computing, through the integration of artificial intelligence (AI) techniques, have facilitated the development of intelligent cognitive systems (ICS). This benefits railway defect detection by enabling ICS to emulate human-like analysis of defect patterns in image data. Although visual defect classification based on convolutional neural networks (CNN) has achieved decent performance, the scarcity of large datasets for railway defect detection remains a challenge. This scarcity stems from the infrequent nature of accidents that result in defective railway parts. Existing research efforts have addressed the challenge of data scarcity by exploring rule-based and generative data augmentation approaches. Among these approaches, variational autoencoder (VAE) models can generate realistic data without the need for extensive baseline datasets for noise modeling. This study proposes a VAE-based synthetic image generation technique for training railway defect classifiers. Our approach introduces a modified regularization strategy that combines weight decay with reconstruction loss. Using this method, we created a synthetic dataset for the Canadian Pacific Railway (CPR), consisting of 50 real samples across five classes. Remarkably, our method generated 500 synthetic samples, achieving a minimal reconstruction loss of 0.021. A visual transformer (ViT) model, fine-tuned using this synthetic CPR dataset, achieved high accuracy rates (98–99%) in classifying the five railway defect classes. This research presents an approach that addresses the data scarcity issue in railway defect detection, indicating a path toward enhancing the development of ICS in this field.
Similar content being viewed by others
Data Availability
We have signed a business agreement with the data provider. Accordingly, the datasets analyzed during the current study are not publicly available. However, we have consent for publishing the result.
References
Zheng Y, Tuan LA, Novel A. Cognitively Inspired, Unified Graph-based Multi-Task Framework for Information Extraction. Cogn Comput. 2023;15:2004–13. https://doi.org/10.1007/s12559-023-10163-2
Gudivada VN, Pankanti S, Seetharaman G, Zhang Y. Cognitive computing systems: Their potential and the future. Comp. 2019;52(5):13–8.
Tabernik D, Šela S, Skvarč J, et al. Segmentation-based deep-learning approach for surface-defect detection. J Intell Manuf. 2020;31:759–76. https://doi.org/10.1007/s10845-019-01476-x.
Tabernik D, Sela S, Skvarˇc J, Skoˇcaj D. Deep-learning-based com-puter vision system for surface-defect detection. In: In Computer Vision Sys-tems: 12th International Conference, ICVS 2019, Thessaloniki, Greece, September 23–25, 2019, Proceedings 12. Springer; 2019. p. 490–500.
Ghaboura S, Ferdousi R, Laamarti F, Yang C, El Saddik A. Digital twin for railway: A comprehensive survey. IEEE Access. 2023;11:120237–57.
Ferdousi R, Laamarti F, Yang C, El Saddik A. Railtwin: a digital twin framework for railway. In 2022 IEEE 18th International Conference on Automation Science and Engineering (CASE). IEEE; 2022. p. 1767–72.
Yang C, Ferdousi R, El Saddik A, Li Y, Liu Z, Liao M. Lifetime learning-enabled modelling framework for digital twin. In: In 2022 IEEE 18th International Conference on Automa-tion Science and Engineering (CASE). 2022. p. 1761–6.
Cui S, Wang H, Zhang M, Zhang X. Defect classification on lim-ited labeled samples with multiscale. Appl Intell. 2021;51(6):3911–25.
Alqudah R, Al-Mousa AA, Hashyeh YA, Alzaibaq OZ. A systemic comparison between using augmented data and syn-thetic data as means of enhancing wafermap defect classification. Comput Ind. 2023;145:103809.
Xiao Y, Huang Y, Li C, et al. Lightweight Multi-modal Representation Learning for RGB Salient Object Detection. Cogn Comput. 2023;15:1868–83. https://doi.org/10.1007/s12559-023-10148-1.
Abufadda M, Mansour K. A survey of synthetic data generation for machine learning. In: In 2021 22nd international arab conference on information technology (ACIT). 2021. p. 1–7.
Jain S, Seth G, Paruthi A, et al. Synthetic data augmentation for surface defect detection and classification using deep learning. J Intell Manuf. 2022;33:1007–20. https://doi.org/10.1007/s10845-020-01710-x.
Zhang G, Cui K, Hung TY, Lu S. Defect-gan: High-fidelity defect synthesis for automated defect inspection. In: In Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision. 2021. p. 2524–34.
Tevosyan A, Khondkaryan L, Khachatrian H, Tade-vosyan G, Apresyan L, Babayan N, Stopper H, Navoyan Z. Improving vae based molecular representations for compound property pre-diction. Journal of Cheminformatics. 2022;14(1):69.
He X, Chang Z, Zhang L, Xu H, Chen H, Luo Z. A survey of de-fect detection applications based on generative adversarial networks. IEEE Access. 2022;10:113493–512.
Lu Y, Shen M, Wang H, Wang X, van Rechem C, Wei W. Machine learning for synthetic data generation: a review. arXiv preprint arXiv:2302.04062, 2023.
Endres M, Mannarapotta Venugopal A, Tran TS. Syn-thetic data generation: a comparative study. In: In Proceedings of the 26th International Database Engineered Applications Symposium. 2022. p. 94–102.
Pinheiro Cinelli L, Ara’ujo Marins M, Bar-ros da Silva EA, Lima Netto S. Variational autoencoder. In: In Varia-tional Methods for Machine Learning with Applications to Deep Networks. Springer; 2021. p. 111–49.
Mak HWL, Han R, Yin HHF. Application of variational autoencoder (vae) model and image processing approaches in game design. Sensors. 2023;23(7):3457.
Kumar T, Mileo A, Brennan R, Bendechache M. Image data augmentation approaches: A comprehensive survey. arXiv preprint arXiv:2301.02830, 2023.
Wang R, Hoppe S, Monari E, Huber MF. De-fect transfer gan: Diverse defect synthesis for data augmentation. arXiv preprint arXiv:2302.08366, 2023.
Zhang G, Cui K, Hung TY, Lu S. Defect-gan: High-fidelity defect synthesis for automated defect inspection. arXiv preprint arXiv:2103.15158, 2021.
Jadon A, Kumar S. Leveraging generative ai models for synthetic data generation in healthcare: Balancing research and privacy. In: In 2023 International Conference on Smart Applications, Communications and Networking (SmartNets). IEEE; 2023. p. 1–4.
Wang Z, Healy G, Smeaton AF, Ward TE. Use of neural signals to evaluate the quality of generative adversarial network performance in facial image generation. Cogn Comp. 2020;12:13–24.
Alpaydin E. Introduction to machine learning. MIT press, 2020.
Shang H, Sun C, Liu J, Chen X, Yan R. Defect-aware transformer network for intelligent visual surface defect detection. Adv Eng Inform. 2023;55: 101882.
Li J, Li D, Savarese S, Hoi S. Blip-2: Boot-strapping language-image pre-training with frozen image encoders and large language models. In: In International conference on machine learning. PMLR; 2023. p. 19730–42.
Funding
The authors extend their appreciation to the researchers supporting project number (RSP2024R32), King Saud University, Riyadh, Saudi Arabia. This research is also supported in part by collaborative research funding from the National Program Office under the National Research Council of Canada’s Artificial Intelligence for Logistics Program. The project ID is AI4L-123.
Author information
Authors and Affiliations
Contributions
Rahatara Ferdousi: Conceptualization, Methodology, Writing—Original Draft Preparation, Visualization, Validation, Software. Chunsheng Yang: Review & Data Collection, Validation, Project administration. M. Anwar Hossain: Writing—Review & Editing, Validation, Supervision. Fedwa Laamarti: Writing—Review & Editing. Abdulmotaleb El Saddik & M. Shamim Hossain: Discussion, Review, Funding acquisition, Supervision.
Corresponding authors
Ethics declarations
Competing interests
The authors declare that there are no conflicts of interest regarding the publication of this paper.
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Supplementary Information
Below is the link to the electronic supplementary material.
Supplementary file1 (MP4 126127 KB)
Rights and permissions
Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
About this article
Cite this article
Ferdousi, R., Yang, C., Hossain, M.A. et al. Generative Model-Driven Synthetic Training Image Generation: An Approach to Cognition in Railway Defect Detection. Cogn Comput (2024). https://doi.org/10.1007/s12559-024-10283-3
Received:
Accepted:
Published:
DOI: https://doi.org/10.1007/s12559-024-10283-3