Attention-Based Multi-Scale Deep Learning for Tomato Leaf Disease Detection in Complex Natural Environments

The present study demonstrates the potential of combining attention mechanisms and multi-scale feature fusion for automated tomato leaf disease detection in complex natural environments. The integration of CBAM into YOLOv6 improves the network's ability to focus on disease-related spatial and channel features while reducing the influence of complex backgrounds. The proposed BiRepGFPN module complements this improvement by integrating shallow spatial information with deeper semantic information and thereby enhancing the detection of small lesions. The complete improved YOLOv6 model achieved 92.9% precision, 95.2% recall, 94.0% F1-score, and 93.8% mAP on the tomato disease dataset. Its performance exceeded the baseline YOLOv6 model by 2.3 percentage points in precision, 4.0 percentage points in recall, 3.1 percentage points in F1-score, and 2.7 percentage points in mAP. The model also demonstrated improved generalization on PlantDoc. Overall, attention-based feature selection and multi-scale feature fusion represent effective strategies for overcoming two major challenges in field-based plant disease detection: environmental interference and small disease lesions. Such approaches can contribute to the development of intelligent agricultural monitoring systems capable of supporting rapid and reliable disease diagnosis. Nevertheless, further work is required to improve robustness across different crops, geographical regions, environmental conditions, disease stages, and image-acquisition systems.

Albattah, W., Nawaz, M., Javed, A., Masood, M., & Albahli, S. (2021). A novel deep learning method for detection and classification of plant diseases. Complex & Intelligent Systems, 8, 507–524. https://doi.org/10.1007/s40747-021-00536-1 Cai, H., & Jiang, J. (2023). An improved plant disease detection method based on YOLOv5. In 2023 15th International Conference on Intelligent Human-Machine Systems and Cybernetics (pp. 237–240). IEEE. https://doi.org/10.1109/IHMSC58761.2023.00062 Ding, X., Zhang, X., Ma, N., Han, J., Ding, G., & Sun, J. (2021). RepVGG: Making VGG-style ConvNets great again. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (pp. 13733–13742). IEEE. https://doi.org/10.1109/CVPR46437.2021.01352 Ge, Z., Liu, S., Wang, F., Li, Z., & Sun, J. (2021). YOLOX: Exceeding YOLO series in 2021. arXiv. https://doi.org/10.48550/arXiv.2107.08430 Guo, W., Feng, Q., Li, X., Yang, S., & Yang, J. (2022). Grape leaf disease detection based on attention mechanisms. International Journal of Agricultural and Biological Engineering, 15, 205–212. https://doi.org/10.25165/j.ijabe.20221505.7548 Hou, Q., Zhou, D., & Feng, J. (2021). Coordinate attention for efficient mobile network design. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (pp. 13708–13717). IEEE. https://doi.org/10.1109/CVPR46437.2021.01350 Jing, J., Li, S., Qiao, C., Li, K., Zhu, X., & Zhang, L. (2023). A tomato disease identification method based on leaf image automatic labeling algorithm and improved YOLOv5 model. Journal of the Science of Food and Agriculture, 103, 7070–7082. https://doi.org/10.1002/jsfa.12793 Li, C., Li, L., Jiang, H., Weng, K., Geng, Y., Li, L., et al. (2022). YOLOv6: A single-stage object detection framework for industrial applications. arXiv. https://doi.org/10.48550/arXiv.2209.02976 Li, S., Li, K., Qiao, Y., & Zhang, L. (2022). A multi-scale cucumber disease detection method in natural scenes based on YOLOv5. Computers and Electronics in Agriculture, 202, 107363. https://doi.org/10.1016/j.compag.2022.107363 Liu, J., & Wang, X. (2021). Plant diseases and pests detection based on deep learning: A review. Plant Methods, 17, 1–18. https://doi.org/10.1186/s13007-021-00722-9 Liu, J., & Wang, X. (2023). Tomato disease object detection method combining prior knowledge attention mechanism and multiscale features. Frontiers in Plant Science, 14. https://doi.org/10.3389/fpls.2023.1255119 Qi, J., Liu, X., Liu, K., Xu, F., Guo, H., Tian, X., et al. (2022). An improved YOLOv5 model based on visual attention mechanism: Application to recognition of tomato virus disease. Computers and Electronics in Agriculture, 194, 106780. https://doi.org/10.1016/j.compag.2022.106780 Singh, D., Jain, N., Jain, P., Kayal, P., Kumawat, S., & Batra, N. (2020). PlantDoc: A dataset for visual plant disease detection. In Proceedings of the 7th ACM IKDD CoDS and 25th COMAD (pp. 249–253). ACM. https://doi.org/10.1145/3371158.3371196 Wang, Y., Zhang, P., & Tian, S. (2024). Tomato leaf disease detection based on attention mechanism and multi-scale feature fusion. Frontiers in Plant Science, 15, 1382802. https://doi.org/10.3389/fpls.2024.1382802