Classification of Flooding in Paddy Field Using Hard Attention-based Deep Learning Model

Jaewon Sok; Kwihoon Kim; Maga Kim; Jin-Yong Choi

doi:10.5389/KSAE.2026.68.5.015

All Issue

2026 Vol.68, Issue 5 Preview Page Next Page

Classification of Flooding in Paddy Field Using Hard Attention-based Deep Learning Model 하드 어텐션 기반 딥러닝 모델을 활용한 논벼 담수 분류

30 September 2026. pp. 15-27

PDF XML

Abstract

Flooding irrigation in the paddy field creates an anaerobic environment in the soil, generating methane emissions. Monitoring flooding irrigation is important to estimate methane emission; however, monitoring flooding depth in the paddy field remains challenging due to technical limitations caused by varying environmental conditions and a lack of cost-effective measurements. This study aimed to develop a deep learning framework incorporating Hard Attention mechanisms to classify flooding or dry conditions in paddy fields. The framework combines three components: a segmentation model that identifies water regions, a base model that analyzes entire images, and a focus model that examines the segmented water surface. These components are integrated through a decision model that combines both local and global features. This study evaluated sequential training (pre-training base and focus models independently) versus simultaneous end-to-end training. The sequentially trained decision model achieved the highest performance (accuracy: 0.85, F1-score: 0.87), while the base model showed inferior results (accuracy: 0.79, F1-score: 0.81), and the focus model demonstrated intermediate performance (accuracy: 0.80, F1-score: 0.85). These findings confirm that integrating global contextual information with targeted water-region features through attention mechanisms enhances classification performance for intermittent irrigation monitoring. Future work will examine model generalizability across diverse environmental conditions.

Keywords

Intermittent irrigation

paddy field

hard attention

deep learning

image segmentation

image classification

References

Archana, R., and P. Jeevaraj, 2024. Deep learning models for digital image processing: A review. Artificial Intelligence Review 57: 1-33. doi:10.1007/s10462-023-10631-z.

10.1007/s10462-023-10631-z

Ba, J., V. Mnih, and K. Kavukcuoglu, 2014. Multiple object recognition with visual attention. arXiv preprint doi:arXiv:1412.7755.

Chen, L.-C., Y. Zhu, G. Papandreou, F. Schroff, and H. Adam, 2018. Encoder-Decoder with Atrous Separable Convolution for Semantic Image Segmentation. In Proc. Computer Vision – ECCV 2018, 833-851. Munich, Germany: Springer-Verlag. doi:10.1007/978-3-030-01234-2_49.

10.1007/978-3-030-01234-2_49

Deng, J., W. Dong, R. Socher, L. J. Li, K. Li, and L. Fei-Fei, 2009. Imagenet: A large-scale hierarchical image database. In Proc. 2009 IEEE Conference on Computer Vision and Pattern Recognition, 248-255. Miami, Fla.: IEEE. doi:10.1109/CVPR.2009.5206848.

10.1109/CVPR.2009.5206848

Elsayed, G. F., S. Kornblith, and Q. V. Le, 2019. Saccader: Improving accuracy of hard attention models for vision. In Proc. 33rd Conference on Neural Information Processing Systems, 700-712. Vancouver, BC, Canada: Curran Associates, Inc.

Fu, J., H. Zheng, and T. Mei, 2017. Look closer to see better: Recurrent attention convolutional neural network for fine-grained image recognition. In Proc. 2017 IEEE Conference on Computer Vision and Pattern Recognition, 4476-4484. Honolulu, Hawaii: IEEE. doi:10.1109/CVPR.2017.476.

10.1109/CVPR.2017.476

Gregor, K., I. Danihelka, A. Graves, D. Rezende, and D. Wierstra, 2015. Draw: A recurrent neural network for image generation. In Proc. The 32nd International Conference on Machine Learning 37: 1462-1471. Lille, France: PMLR.

Gwon, H. S., E. J. Choi, S. I. Lee, H. S. Lee, J. M. Lee, and S. S. Kang, 2022. Research review of methane emissions from korean rice paddies. Journal of Climate Change Research 13(1): 117-134 (in Korean). doi:10.15531/KSCCR.2022.13.1.117.

10.15531/KSCCR.2022.13.1.117

Hedderich, R., and W. B. Whitman, 2006. Physiology and biochemistry of the methane-producing Archaea. The Prokaryotes 2: 1050-1079. doi:10.1007/0-387-30742-7_34.

10.1007/0-387-30742-7_34

Howard, A., M. Sandler, G. Chu, L.-C. Chen, B. Chen, M. Tan, W. Wang, Y. Zhu, R. Pang, V. Vasudevan, Q. V. Le, and H. Adam, 2019. Searching for MobileNetV3. In Proc. 2019 IEEE/CVF International Conference on Computer Vision, 1314-1324. Seoul, Korea: IEEE. doi:10.1109/ICCV.2019.00140.

10.1109/ICCV.2019.00140

Hu, J., L. Shen, and G. Sun, 2018. Squeeze-and-excitation networks. In Proc. The IEEE Conference on Computer Vision and Pattern Recognition, 7132-7141. Salt Lake City, Utah.: IEEE. doi:10.1109/CVPR.2018.00745.

10.1109/CVPR.2018.00745

Khan, A., A. Sohail, U. Zahoora, and A. S. Qureshi, 2020. A survey of the recent architectures of deep convolutional neural networks. Artificial Intelligence Review 53(8): 5455-5516.

10.1007/s10462-020-09825-6

Kingma, D. P., and J. Ba, 2014. Adam: A method for stochastic optimization. arXiv preprint doi:arXiv:1412.6980.

LeCun, Y., Y. Bengio, and G. Hinton, 2015. Deep learning. Nature 521(7553): 436-444.

10.1038/nature14539

MAFRA (Ministry of Agriculture, Food and Rural Affairs), 2025. Recruitment opens for the 'Low-Carbon Agriculture Program' targeting rice farmers. https://www.korea.kr/briefing/pressReleaseView.do?newsId=156679605. Accessed 19 Mar. 2025 (in Korean).

Mnih, V., N. Heess, A. Graves, and K. Kavukcuoglu, 2014. Recurrent models of visual attention. In Proc. 28th International Conference on Neural Information Processing Systems, 2204-2212. Montreal, Canada: MIT Press.

Sandler, M., A. Howard, M. Zhu, A. Zhmoginov, and L. C. Chen, 2018. Mobilenetv2: Inverted residuals and linear bottlenecks. In Proc. The IEEE Conference on Computer Vision and Pattern Recognition, 4510-4520. Salt Lake City, Utah.: IEEE. doi:10.1109/CVPR.2018.00474.

10.1109/CVPR.2018.00474

Shin, H., J. Lee, T. Eo, Y. Jun, S. Kim, and D. Hwang, 2020. The latest trends in attention mechanisms and their application in medical imaging. Journal of the Korean Society of Radiology 81(6): 1305-1333 (in Korean). doi:10.3348/jksr.2020.0150.

10.3348/jksr.2020.0150

Shrestha, A., and A. Mahmood, 2019. Review of deep learning algorithms and architectures. IEEE Access 7, 53040-53065.

10.1109/ACCESS.2019.2912200

Toma, Y., Y. Takechi, A. Inoue, N. Nakaya, K. Hosoya, Y. Yamashita, and U. Hideto, 2021. Early mid-season drainage can mitigate greenhouse gas emission from organic rice farming with green manure application. Soil Science and Plant Nutrition 67(4): 482-492. doi:10.1080/00380768.2021.1927832.

10.1080/00380768.2021.1927832

Wandell, B. A., 1995. Foundations of Vision. Sunderland, Mass.: Sinauer Associates.

Wang, D., W. Cao, F. Zhang, Z. Li, S. Xu, and X. Wu, 2022. A review of deep learning in multiscale agricultural sensing. Remote Sensing 14(3): 559.

10.3390/rs14030559

Wang, P., E. Fan, and P. Wang, 2020. Comparative analysis of image classification algorithms based on traditional machine learning and deep learning. Pattern Recognition Letters 141: 61-67. doi:10.1016/j.patrec.2020.07.042.

10.1016/j.patrec.2020.07.042

Information

Publisher :The Korean Society of Agricultural Engineers
Publisher(Ko) :한국농공학회
Journal Title :Journal of Korean Society of Agricultural Engineers
Journal Title(Ko) :한국농공학회논문집
Volume : 68
No :5
Pages :15-27
Received Date : 2025-09-01
Revised Date : 2026-04-15
Accepted Date : 2026-05-26
DOI :https://doi.org/10.5389/KSAE.2026.68.5.015

Journal of Korean Society of Agricultural Engineers ISSN:1738-3692(Print) 2093-7709(Online) 한국농공학회논문집

All Issue