Data augmentation method for insulators based on Cycle GAN

Abstract

Data augmentation is an important task of using existing data to expand data sets. Using generative countermeasure network technology to realize data augmentation has the advantages of high-quality generated samples, simple training, and fewer restrictions on the number of generated samples. However, in the field of transmission line insulator images, the freely synthesized samples are prone to produce fuzzy backgrounds and disordered samples of the main insulator features. To solve the above problems, this paper uses cycle generative adversarial network (Cycle-GAN) used for domain conversion in the generation countermeasure network as the initial framework and uses the self-attention mechanism and the channel attention mechanism to assist the conversion to realize the mutual conversion of different insulator samples. The attention module with prior knowledge is used to build the generation countermeasure network, and the GAN model with local controllable generation is built to realize the directional generation of insulator belt defect samples. The experimental results show that the samples obtained by this method are improved in a number of quality indicators, and the quality effect of the samples obtained is excellent, which has a reference value for the data expansion of insulator images.

Keywords

Data expansion / Deep learning / Generate confrontation network / Insulator

Funding

This work was supported in part by the National Natural Science Foundation of China under Grant No. 61973055, Fundamental Research Funds for the Central Universities under Grant No. ZYGX2020J011, and Regional Innovation Cooperation Funds of Sichuan under Grant No. 2024YFHZ0089.

Author contributions

R. Ye and X.-S. Yu contributed to the idea of this article; C. Zhang and X.-S. Yu performed the experiments; R. Ye and X.-S. Yu wrote the manuscript; A. Boukerche, B. Yan, and X.-J. Zhou contributed to the revision of this paper.

Declaration of competing interest

The author declare no conflicts of interest.

References

[1] Y. Tokozume, Y. Ushiku, T. Harada, Learning from between- class examples for deep sound recognition, in: Proc. of the 6th Intl. Conf. on Learning Representations, Vancouver, Canada, (2018), pp. 1-13.

[2] R. Takahashi, T. Matsubara, K. Uehara RICAP, Random image cropping and patching data augmentation for deepCNNs, in: Proc. of the 10th Asian Conf. on Machine Learning, Beijing, China, (2018), pp. 786-798.

[3] I.J. Goodfellow, J. Pouget-Abadie, M. Mirza, et al., Generative adversarial nets, in: Proc. of the 27th Intl. Conf. on Neural Information Processing Systems, Montreal, Canada, (2014), pp. 2672-2680.

[4] A. Antoniou, A. Storkey, H. Edwards, Data Augmentation Generative Adversarial Networks, https://arxiv.org/abs/1711.04340, November 2017.

[5] B.M. Lake, R. Salakhutdinov, J.B. Tenenbaum, Human-level concept learning through probabilistic program induction, Science, 350 (6266) (Dec. 2015), pp. 1332-1338.

[6] A. Radford, L. Metz, S. Chintala, Unsupervised representation learning with deep convolutional generative adversarial networks, in: Proc. of the 4th Intl. Conf. on Learning Representations, San Juan, America, (2016), pp. 1-15.

[7] M. Frid-Adar, E. Klang, M. Amitai, J. Goldberger, H. Greenspan, Synthetic data augmentation using GAN for improvedliver lesion classification, in: Proc. of the 15th IEEE Intl. Symposium on Biomedical Imaging, Washington, America, (2018), pp. 289-293.

[8] C. Han, K. Murao, T. Noguchi, et al., Learning more with less: conditional PGGAN-based data augmentation for brain metastases detection using highly-rough annotation on MR images, in: Proc. of the 28th ACM International Conference on Information and Knowledge Management (CIKM’19), Association for Computing Machinery, Beijing, China, (2019), pp. 119-127.

[9] J. Redmon, A. Farhadi, YOLOv3: an Incremental Improvement [Online]. Available, https://arxiv.org/abs/1804.02767, April 2018.

[10] X.-Y. Zhu, Y.-F. Liu, J.-H. Li, T. Wan, Z.-C. Qin, Emotion classification with data augmentation using generative adversarial network, in: Proc. of the 22nd Pacific-Asia Conf. on Knowledge Discovery and Data Mining, Melbourne, Australia, (2018), pp. 349-360.

[11] J.-Y. Zhu, T. Park, P. Isola, A.A. Efros, Unpaired image-to-image translation using cycle-consistent adversarial networks, in: Proc. of IEEE Intl. Conf. on Computer Vision, Venice, Italy, (2017), pp. 2242-2251.

[12] T. Kim, M. Cha, H. Kim, J.K. Lee, J. Kim, Learning to discover cross-domain relations with generative adversarial networks, in: Proc. of the 34th Intl. Conf. on Machine Learning, Sydney, Australia, (2017), pp. 1857-1865.

[13] Z.-L. Yi, H. Zhang, P. Tan, M.-L. Gong, DualGAN: unsupervised dual learning for image-to-image translation, in: Proc. of IEEE Intl. Conf. on Computer Vision, Venice, Italy, (2017), pp. 2868-2876.

[14] J. Hu, L. Shen, S. Albanie, G. Sun, E.-H. Wu, Squeeze- and- excitation networks, IEEE T. Pattern Anal. 42 (8) (Aug 2020) 2011-2023.

[15] Z. Liang, J.-X. Huang, CycleGAN with dynamic criterion for malaria blood cell image synthetization, AMIA J. Summ. Transl. Sci. Proc., (2022), pp. 323-330.

[16] S. Benaim, L. Wolf, One-sided unsupervised domain mapping, in: Proc. of the 31st Intl. Conf. on Neural Information Processing Systems, Long Beach, America, (2017), pp. 752-762.

[17] B. Liu, J. Lv, X. Fan, et al., Application of an improved DCGAN for image generation, Mobile Inf. Syst. 2022 (Pt.16) (2022) 1-14.

[18] M. Arjovsky, S. Chintala, L. Bottou, Wasserstein generative adversarial networks, in: Proc. of the 34th Intl. Conf. on Machine Learning, Sydney, Australia, (2017), pp. 214-223.

[19] I. Gulrajani, F. Ahmed, M. Arjovsky, V. Dumoulin, A. Courville, Improved training of Wasserstein GANs, in: Proc. of the 31st Intl. Conf. on Neural Information Processing Systems, Long Beach, America, (2017), pp. 5769-5779.