Cassava leaf disease image recognition method for imbalanced data

doi:10.13733/j.jcam.issn.2095-5553.2025.03.016

Abstract

Abstract:

To improve the accuracy of automatic cassava leaf disease recognition in production environments and address issues such as low-contrast disease images and long-tail data distribution, this paper proposes a deep learning model, SwinTFCC, for the recognition of cassava leaf diseases. This model employs the Swin Transformer as the backbone network, utilizing its self-attention mechanism and hierarchical structure to focus on local and global features for robust disease recognition in complex backgrounds. The features from the last layer are then input into a feature cluster compression module to map sparse feature clusters into dense ones, reducing classification errors caused by sparse feature clusters of underrepresented classes crossing decision boundaries in the long-tailed distribution. The model is trained on a cassava leaf disease image dataset using transfer learning to enhance recognition performance. The experimental results indicate that the proposed model achieves an F1 score of 90.74%, improving by 8.04% to 19.3% compared with other mainstream models. In this study, the method performs well on a small-scale imbalanced dataset, confirming the model's effectiveness and providing technical support for the automatic and precise recognition of cassava leaf diseases.

Key words: cassava leaf, disease recognition, image recognition, transfer learning, imbalanced data

摘要：

为提高产地环境下木薯叶部病害自动识别的准确性，解决病害图像低对比度和数据长尾分布问题，建立一种深度学习模型SwinTFCC用于木薯叶部病害识别。该模型采用Swin Transformer作为骨干网络，借助Swin Transformer的自注意力机制和层级结构关注局部与全局特征，使其对复杂背景病害识别具有鲁棒性；将最后一层特征输入特征簇压缩模块，以映射稀疏特征簇为稠密特征簇，减少长尾分布中样本少的类别稀疏特征簇跨越决策边界导致分类错误情况；并采用迁移学习在木薯叶部病害图像数据集上进行训练，以提升木薯叶部病害识别性能。试验结果表明，模型的F1值达到90.74%，较其他主流模型提升8.04%~19.3%。所采用的方法在小规模不平衡数据集上取得较好效果，验证模型的有效性，为木薯叶部病害自动精准识别提供技术支撑。

关键词: 木薯叶部, 病害识别, 图像识别, 迁移学习, 不平衡数据

CLC Number:

Wang Danyang, Liang Weihong, Li Yuping, Huang Guixiu, . Cassava leaf disease image recognition method for imbalanced data[J]. Journal of Chinese Agricultural Mechanization, 2025, 46(3): 101-107.

王丹阳, 梁伟红, 李玉萍, 黄贵修, . 面向不平衡数据的木薯叶部病害图像识别方法[J]. 中国农机化学报, 2025, 46(3): 101-107.

References

［1］　FAO.The state of food security and nutrition in the world 2023 ［R］. Rome: FAO, 2023.
［2］　李丛希, 谭砚文. 新冠肺炎疫情下世界木薯产业发展趋势、挑战及对中国的启示［J］. 农业展望, 2022, 18（10）: 26-32.Li Congxi, Tan Yanwen.Development trends, challenges and enlightenment to China of the world cassava industry under the COVID-19 pandemic ［J］. Agricultural Outlook, 2022, 18（10）: 26-32.
［3］　胡敏, 陈红波, 许良凤, 等.基于颜色和纹理特征的黄瓜病害识别算法［J］. 电子测量与仪器学报, 2015, 29（7）: 970-977.〖JP2〗Hu Min, Chen Hongbo, Xu Liangfeng, et al. Cucumber disease recognition algorithm based on color and texture features ［J］. Journal of Electronic Measurement and Instrumentation, 2015, 29（7）: 970-977.
［4］　胡维炜, 张武, 刘连忠.基于Variance—SFFS的小麦叶部病害图像识别［J］. 湖南农业大学学报（自然科学版）, 2018, 44（2）: 225-228.Hu Weiwei, Zhang Wu, Liu Lianzhong. Identification of wheat leaf diseases based on Variance—SFFS algorithm ［J］. Journal of Hunan Agricultural University （Natural Sciences）, 2018, 44（2）: 225-228.
［5］　Parikh A, Raval M S, Parmar C, et al. Disease detection and severity estimation in cotton plant from unconstrained images ［C］. 2016 IEEE International Conference on Data Science and Advanced Analytics （DSAA）, 2016: 594-601.
［6］　Suresha M, Shreekanth K, Thirumalesh B. Recognition of diseases in paddy leaves using kNN classifier ［C］. 2017 2nd International Conference for Convergence in Technology （I2CT）, 2017: 663-666.
［7］　翟治芬, 徐哲, 周新群, 等. 基于朴素贝叶斯分类器的棉花盲椿象危害等级识别［J］. 农业工程学报, 2015, 31（1）: 204-211.
Zhai Zhifen, Xu Zhe, Zhou Xinqun, et al. Recognition of hazard grade for cotton blind stinkbug based on Naive Bayesian classifier ［J］. Transactions of the Chinese Society of Agricultural Engineering, 2015, 31（1）: 204-211.
［8］　Xiao T, Liu H, Cheng Y. Corn disease identification based on improved GBDT method ［C］. 2019 6th International Conference on Information Science and Control Engineering （ICISCE）, 2019: 215-219.
［9］　Sladojevic S, Arsenovic M, Anderla A, et al. Deep neural networks based recognition of plant diseases by leaf image classification ［J］. Computational Intelligence and Neuroscience, 2016.
［10］　Agarwal M, Gupta S K, Biswas K.Development of Efficient CNN model for tomato crop disease identification ［J］. Sustainable Computing: Informatics and Systems, 2020, 28: 100407.
［11］　Wu Zhao, Jiang Feng, Cao Rui. Research on recognition method of leaf diseases of woody fruit plants based on transfer learning ［J］. Scientific Reports, 2022, 12（1）: 15385.
［12］　卫雅娜, 王志彬, 乔晓军, 等. 基于注意力机制与EfficientNet的轻量化水稻病害识别方法［J］. 中国农机化学报, 2022, 43（11）: 172-181.
Wei Yana, Wang Zhibin, Qiao Xiaojun, et al. Lightweight rice disease identification method based on attention mechanism and EfficientNet ［J］. Journal of Chinese Agricultural Mechanization, 2022, 43（11）: 172-181.
［13］　Haque M A, Marwaha S, Deb C K, et al.Deep learningbased approach for identification of diseases of maize crop ［J］. Scientific Reports, 2022, 12（1）: 6334.
［14］　Li Gaoqiang, Jiao Lin, Chen Peng, et al. Spatial convolutional selfattentionbased transformer module for strawberry disease identification under complex background ［J］. Computers and Electronics in Agriculture, 2023, 212: 108121.
［15］　Dosovitskiy A, Beyer L, Kolesnikov A, et al. An image is worth 16×16 words: Transformers for image recognition at scale ［C］. International Conference on Learning Representations, 2021.

［16］　Han Kai, Xiao An, Wu Enhua, et al. Transformer in transformer ［J］. Advances in Neural Information Processing Systems, 2021, 34: 15908-15919.

［17］　Touvron H, Cord M, Douze M, et al. Training dataefficient image transformers & distillation through attention ［C］. Proceedings of the 38th International Conference on Machine Learning. PMLR, 2021: 10347-10357.
［18］　Liu Z, Lin Y, Cao Y, et al. Swin Transformer: Hierarchical vision transformer using shifted windows ［C］. 2021 IEEE/CVF International Conference on Computer Vision （ICCV）, 2021: 9992-10002.
［19］　Chawla N V, Japkowicz N, Kotcz A. Special issue on learning from imbalanced data sets ［J］. ACM SIGKDD Explorations Newsletter, 2004, 6（1）: 1-6.
［20］　Mani I, Zhang I. Knn approach to unbalanced data distributions: A case study involving information extraction ［C］. Proceedings of Workshop on Learning from Imbalanced Datasets, 2003: 1-7.
［21］　Kang B, Xie S, Rohrbach M, et al. Decoupling representation and classifier for longtailed recognition ［J］. arXiv:191009217, 2019.
［22］　Lin T Y, Goyal P, Girshick R, et al. Focal loss for dense object detection ［J］. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2020, 42（2）: 318-327.
［23］　Szegedy C, Vanhoucke V, Ioffe S, et al. Rethinking the inception architecture for computer vision ［C］. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2016: 2818-2826.
［24］　Li Jian, Meng Ziyao, Shi Daqian, et al. FCC: Feature clusters compression for longtailed visual recognition ［C］. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2023: 24080-24089.
［25］　Li X, Chen X, Yang J, et al. Transformer helps identify kiwifruit diseases in complex natural environments ［J］. Computers and Electronics in Agriculture, 2022, 200: 107258.
［26］　Wang Fengyi, Rao Yuan, Luo Qing, et al. Practical cucumber leaf disease recognition using improved Swin Transformer and small sample size ［J］. Computers and Electronics in Agriculture, 2022, 199: 107163.

[1]	Li Chunyu, Guo Xiaoqin, Yang Jingjing, Li Zhonghua. Fruit image recognition based on deep learning [J]. Journal of Chinese Agricultural Mechanization, 2025, 46(1): 198-203.
[2]	Niu Panting, Zhang Baolin, , , Pan Lijie, Guo Jianpeng. Identification and explainability of maize leaf diseases based on improved GoogLeNet#br# [J]. Journal of Chinese Agricultural Mechanization, 2025, 46(1): 204-212.
[3]	Zhao Shuang, Yu Yongqiang, Miao Yubin, Liu Kexin. Research of impurity detection of green vegetable based on improved Mask R-CNN [J]. Journal of Chinese Agricultural Mechanization, 2024, 45(9): 77-82.
[4]	Han Pengfei, Song Qijiang, Jia Mengshi. Wheat seed classification based on improved lightweight EfficientNet-V2 model [J]. Journal of Chinese Agricultural Mechanization, 2024, 45(9): 111-117.
[5]	Song Xiaolu, Feng Juan, Liang Xiangyu, Zhou Xixing. Recognition and counting algorithm of underwater sea cucumbers based on YOLOv4 network [J]. Journal of Chinese Agricultural Mechanization, 2024, 45(9): 258-264.
[6]	Su Fei, , Wang Guanghui, Shi Yanxia, Jia Ran, Yan Yinfa, , Zu Linlu, . Leaf disease segmentation model of greenhouse tomatoes based on ResUnet with attention mechanism [J]. Journal of Chinese Agricultural Mechanization, 2024, 45(8): 228-233.
[7]	Huang Yiqi, Lu Linfei, Shen Hao, , Wang Fukuan, , Qiao Xi, . Research on invading insect recognition based on convolutional neural network [J]. Journal of Chinese Agricultural Mechanization, 2024, 45(7): 222-227.
[8]	Zhang Jicheng, Huang Xiangdang. Maize disease identification method based on improved YOLOv3 [J]. Journal of Chinese Agricultural Mechanization, 2024, 45(7): 269-275.
[9]	Li Longjie, Shi Yong, , Liu Yancen, , Guo Junxian, . Research on predicting Hami melon leaf chlorophyll based on RGB image processing [J]. Journal of Chinese Agricultural Mechanization, 2024, 45(6): 149-155.
[10]	Huang Yao, He Jing, Fu Rao, Liu Gang, , Lin Yuanyang. Application analysis of the YOLOv5s-CBAM algorithm for the identification of eggs of Pomacea [J]. Journal of Chinese Agricultural Mechanization, 2024, 45(6): 223-228.
[11]	Shi Fang, Wang Ying, Wang Xinfa, Ma Yukun. Research on potato disease identification based on RegNet network [J]. Journal of Chinese Agricultural Mechanization, 2024, 45(6): 229-234.
[12]	Guo Hongjie, Ma Rui, Wang Jia, Zhao Wei, Ma Dexin. Research on apple leaf disease image recognition based on convolutional neural network [J]. Journal of Chinese Agricultural Mechanization, 2024, 45(5): 239-245.
[13]	Wang Ling, Zhang Jinxiong, Tian Hui, Zhang Zhujun, Zhao Shuhan. Automatic reading of wheat microprecipitation values based on image processing [J]. Journal of Chinese Agricultural Mechanization, 2024, 45(4): 149-154.
[14]	Ma Zhiyan, , Li Hui, Yang Guangyou, . Research on key technologies of intelligent tea picking machine based on YOLOv3 algorithm [J]. Journal of Chinese Agricultural Mechanization, 2024, 45(4): 199-204.
[15]	Qiu Chang, Tian Guangzhao, Zhao Jiawei, Xie Shangjie, Zheng Kui. Strawberry disease identification based on improved YOLOv5 [J]. Journal of Chinese Agricultural Mechanization, 2024, 45(3): 198-204.