Romiyal George
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Romiyal George has a total of 18 co-authors. Here are the top co-authors by number of shared publications:
- Selvarajah Thuseethan
Charles Darwin University
4 shared publications - Roshan Ragel
University of Peradeniya
3 shared publications - Ashen Iranga Hewarathna
Faculty of Computing, Sabaragamuwa University of Sri Lanka, Belihuloya 70140, Sri Lanka
1 shared publication - Bharanidharan Shanmugam
Energy and Resources Institute, Faculty of Science and Technology, Charles Darwin University, Darwin, NT 0810, Australia
1 shared publication Chathrie Wimalasooriya
1 shared publication- Chathrie Wimalasooriya
School of Computing, University of Otago, Dunedin 9016, Otago, New Zealand
1 shared publication
Latest Publications
- Open Access
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- p. 1
Robustness of Lightweight Neural Networks for EEG-based Emotion Recognition
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- p. 1
Open AccessRobustness of Lightweight Neural Networks for EEG-based Emotion Recognition
- Open Access
- 28 July 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 133668-133680
Background-Masked Lightweight Approach for Pear Leaf Disease Recognition
Deep learning for plant disease recognition faces challenges in real-world scenarios, especially on resource-constrained devices and with complex backgrounds. While deep neural networks excel with such images, lightweight approaches struggle due to the complexities in feature extraction, affecting classification accuracy. To overcome this, we introduce a novel pipeline that masks and removes complex backgrounds, enabling accurate disease recognition from affected leaves. This approach also significantly increases model accuracy and reduces training time compared to baseline models due to object-cropped input. Using various lightweight neural networks as backbones for the classification task, we validated the efficiency of our approach, which outperformed the baseline in terms of performance. Specifically, our proposed background masking model achieved an IoU of 93.93% for background masking. Furthermore, for disease recognition, our approach using the SqueezeNet-1.1 backbone demonstrated the highest accuracy of 90.84%, surpassing the baseline’s highest average accuracy of 89.22%. These results demonstrate the efficacy of our proposed method in enhancing disease recognition in agricultural imagery, thereby contributing to the advancement of deep learning techniques in this domain. Moreover, the comparison of training times demonstrates that our approach achieved a considerable reduction compared to the benchmark.
- 28 July 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 133668-133680
Open AccessBackground-Masked Lightweight Approach for Pear Leaf Disease Recognition
Deep learning for plant disease recognition faces challenges in real-world scenarios, especially on resource-constrained devices and with complex backgrounds. While deep neural networks excel with such images, lightweight approaches struggle due to the complexities in feature extraction, affecting classification accuracy. To overcome this, we introduce a novel pipeline that masks and removes complex backgrounds, enabling accurate disease recognition from affected leaves. This approach also significantly increases model accuracy and reduces training time compared to baseline models due to object-cropped input. Using various lightweight neural networks as backbones for the classification task, we validated the efficiency of our approach, which outperformed the baseline in terms of performance. Specifically, our proposed background masking model achieved an IoU of 93.93% for background masking. Furthermore, for disease recognition, our approach using the SqueezeNet-1.1 backbone demonstrated the highest accuracy of 90.84%, surpassing the baseline’s highest average accuracy of 89.22%. These results demonstrate the efficacy of our proposed method in enhancing disease recognition in agricultural imagery, thereby contributing to the advancement of deep learning techniques in this domain. Moreover, the comparison of training times demonstrates that our approach achieved a considerable reduction compared to the benchmark.
- Top Cited PapersOpen Access
- 12 March 2025
- Published by Elsevier in Computers and Electronics in Agriculture
- Vol. 234, 110128
Past, present and future of deep plant leaf disease recognition: A survey
- 12 March 2025
- Published by Elsevier in Computers and Electronics in Agriculture
- Vol. 234, 110128
Top Cited PapersOpen AccessPast, present and future of deep plant leaf disease recognition: A survey
- Open Access
- 12 September 2024
- Published by MDPI AG in Technologies
- Vol. 12 (9) , 160
Change Detection for Forest Ecosystems Using Remote Sensing Images with Siamese Attention U-Net
Forest ecosystems are critical components of Earth’s biodiversity and play vital roles in climate regulation and carbon sequestration. They face increasing threats from deforestation, wildfires, and other anthropogenic activities. Timely detection and monitoring of changes in forest landscapes pose significant challenges for government agencies. To address these challenges, we propose a novel pipeline by refining the U-Net design, including employing two different schemata of early fusion networks and a Siam network architecture capable of processing RGB images specifically designed to identify high-risk areas in forest ecosystems through change detection across different time frames in the same location. It annotates ground truth change maps in such time frames using an encoder–decoder approach with the help of an enhanced feature learning and attention mechanism. Our proposed pipeline, integrated with ResNeSt blocks and SE attention techniques, achieved impressive results in our newly created forest cover change dataset. The evaluation metrics reveal a Dice score of 39.03%, a kappa score of 35.13%, an F1-score of 42.84%, and an overall accuracy of 94.37%. Notably, our approach significantly outperformed multitasking model approaches in the ONERA dataset, boasting a precision of 53.32%, a Dice score of 59.97%, and an overall accuracy of 97.82%. Furthermore, it surpassed multitasking models in the HRSCD dataset, even without utilizing land cover maps, achieving a Dice score of 44.62%, a kappa score of 11.97%, and an overall accuracy of 98.44%. Although the proposed model had a lower F1-score than other methods, other performance metrics highlight its effectiveness in timely detection and forest landscape monitoring, advancing deep learning techniques in this field.
- 12 September 2024
- Published by MDPI AG in Technologies
- Vol. 12 (9) , 160
Open AccessChange Detection for Forest Ecosystems Using Remote Sensing Images with Siamese Attention U-Net
Forest ecosystems are critical components of Earth’s biodiversity and play vital roles in climate regulation and carbon sequestration. They face increasing threats from deforestation, wildfires, and other anthropogenic activities. Timely detection and monitoring of changes in forest landscapes pose significant challenges for government agencies. To address these challenges, we propose a novel pipeline by refining the U-Net design, including employing two different schemata of early fusion networks and a Siam network architecture capable of processing RGB images specifically designed to identify high-risk areas in forest ecosystems through change detection across different time frames in the same location. It annotates ground truth change maps in such time frames using an encoder–decoder approach with the help of an enhanced feature learning and attention mechanism. Our proposed pipeline, integrated with ResNeSt blocks and SE attention techniques, achieved impressive results in our newly created forest cover change dataset. The evaluation metrics reveal a Dice score of 39.03%, a kappa score of 35.13%, an F1-score of 42.84%, and an overall accuracy of 94.37%. Notably, our approach significantly outperformed multitasking model approaches in the ONERA dataset, boasting a precision of 53.32%, a Dice score of 59.97%, and an overall accuracy of 97.82%. Furthermore, it surpassed multitasking models in the HRSCD dataset, even without utilizing land cover maps, achieving a Dice score of 44.62%, a kappa score of 11.97%, and an overall accuracy of 98.44%. Although the proposed model had a lower F1-score than other methods, other performance metrics highlight its effectiveness in timely detection and forest landscape monitoring, advancing deep learning techniques in this field.