Explainable Deep Learning Models for Citrus Leaf
Disease Identification and Agricultural Decision Support
Citrus is a commercially important
crop affected by diseases such as anthracnose, bacterial blight, and citrus
canker, which damage trees and reduce fruit yield. Classical diagnosis is
time-consuming and expert-dependent, limiting scalability. In this paper, we
propose explainable deep learning models for multiclass citrus leaf disease
detection, incorporating custom pre-processing and extensive data augmentation
for robustness. A custom CNN and transfer learning-based VGG16 were fine-tuned
on an augmented dataset of nine citrus leaf categories, including both healthy
and unhealthy specimens. Performance was evaluated using accuracy, precision,
recall, F1-score, and confusion matrices. The CNN achieved 97% accuracy,
outperforming VGG16’s 95%, demonstrating the effectiveness of our preprocessing
approach. For interpretability, Grad-CAM, SHAP, and LIME were uniformly applied
to highlight symptom-relevant regions aligned with agronomic knowledge,
enhancing transparency and user trust. The framework effectively combines
predictability and interpretability, supporting practical, real-time, and
sustainable agricultural decision systems.
[1] P. Sharma and P. Abrol, “Multi-component
image analysis for citrus disease detection using convolutional neural
networks,” Crop Protection, vol. 193, p. 107181, Jul. 2025, doi:
10.1016/j.cropro.2025.107181.
[2] A. Goyal and K. Lakhwani, “Integrating
advanced deep learning techniques for enhanced detection and classification of
citrus leaf and fruit diseases,” Scientific Reports, vol. 15, no. 1, p. 12659,
Apr. 2025, doi: 10.1038/s41598-025-97159-0.
[3] K. Yan et al., “Three-dimensional
fluorescence spectral characteristic of flavonoids for citrus Huanglongbing
disease early detection,” Microchemical Journal, vol. 208, p. 112263, Jan.
2025, doi: 10.1016/j.microc.2024.112263.
[4] S. A. Higgins et al., “Plant-derived,
nodule-specific cysteine-rich peptides as a novel source of biopesticides for
controlling citrus greening disease,” Phytopathology, vol. 114, no. 5, pp.
971–981, May 2024, doi: 10.1094/PHYTO-09-23-0322-KC.
[5] P. K. Yadav et al., “Automated
classification of citrus disease on fruits and leaves using convolutional
neural network generated features from hyperspectral images and machine
learning classifiers,” Journal of Applied Remote Sensing, vol. 18, no. 1, p.
014512, Jan. 2024, doi: 10.1117/1.JRS.18.014512.
[6] P.
Dinesh and R. Lakshmanan, “Deep learning-driven citrus disease detection: A
novel approach with DeepOverlay L-UNet and VGG-RefineNet,” International
Journal of Advanced Computer Science and Applications, vol. 15, no. 7, pp.
1023–1041, Jul. 2024.
[7] C. Wang et al., “A simple and effective
VIGS system facilitates the control of citrus canker by silencing CsLOB1,”
Phytopathology Research, vol. 6, no. 1, p. 12, Mar. 2024, doi:
10.1186/s42483-024-00234-z.
[8] R. Sujatha, S. Krishnan, J. M. Chatterjee,
and A. H. Gandomi, "Advancing plant leaf disease detection integrating
machine learning and deep learning," Scientific Reports, vol. 15, Art.
no. 11552, Apr. 2025, doi: 10.1038/s41598-024-72197-2.
[9] P. Dhiman, A. Kaur, Y. Hamid, E.
Alabdulkreem, H. Elmannai, and N. Ababneh, "Smart Disease Detection
System for Citrus Fruits Using Deep Learning with Edge Computing,"
Sustainability, vol. 15, no. 5, p. 4576, Mar. 2023, doi: 10.3390/su15054576.
[10] H. K. Kondaveeti and C. G. Simhadri,
"Evaluation of deep learning models using explainable AI with qualitative
and quantitative analysis for rice leaf disease detection," Scientific
Reports, vol. 15, Art. no. 31850, Aug. 2025, doi: 10.1038/s41598-025-14306-3.
[11] Q. Xu, L. Ma, R. Zhou, C. Wang, J. Bai, L.
Sun, and J. Cai, “Detection of huanglongbing infection in citrus using
compositional analysis of volatile organic compounds,” Plant Pathology, vol.
73, no. 8, pp. 2084–2100, Oct. 2024, doi: 10.1111/ppa.13964.
[12] X. Guo, X. Wang, W. Zhu, S. X. Yang, L.
Song, P. Li, and Q. Li, “Citrus disease detection based on dilated
reparameterized feature enhancement and shared parameter head,” Sensors, vol.
25, no. 7, p. 1971, Mar. 2025, doi: 10.3390/s25071971.
[13] E. Aswini and C. Vijayakumaran, “Citrus
plant classification using gated fusion adaptive graph neural network with
harbor seal whiskers optimization algorithm,” Knowledge-Based Systems, vol.
329, p. 114093, Nov. 2025, doi: 10.1016/j.knosys.2025.114093.
[14] H. Zhu, D. Wang, Y. Wei, X. Zhang, and L.
Li, "Combining Transfer Learning and Ensemble Algorithms for Improved
Citrus Leaf Disease Classification," Agriculture, vol. 14, no. 9, p.
1549, Sep. 2024. doi: 10.3390/agriculture14091549. [Online]. Available:
https://doi.org/10.3390/agriculture14091549
[15] B.
Kaur, S. K. Gupta, M. Janarthan, D. M. Alsekait, and D. S. AbdElminaam,
"Precision diagnosis of citrus leaf diseases using image enhancement and
nonlinear fuzzy ranking ensemble approach NLFuRBe," Scientific Reports,
vol. 15, Art. no. 32296, 2025. doi: 10.1038/s41598-025-16923-4. [Online].
Available: https://doi.org/10.1038/s41598-025-16923-4
[16] H.
Mo and L. Wei, "Lightweight citrus leaf disease detection model based on
ARMS and cross-domain dynamic attention," Journal of King Saud University
– Computer and Information Sciences, vol. 36, no. 10, p. 102133, Oct. 2024.
doi: 10.1016/j.jksuci.2024.102133. [Online]. Available:
https://doi.org/10.1016/j.jksuci.2024.102133
[17] H. J. Min, J. Qin, P. K. Yadav, Q.
Frederick, T. Burks, M. Dewdney, I. Baek, and M. Kim, "Classification of
Citrus Leaf Diseases Using Hyperspectral Reflectance and Fluorescence Imaging
and Machine Learning Techniques," Horticulturae, vol. 10, no. 11, p.
1124, Oct. 2024. doi: 10.3390/horticulturae10111124. [Online]. Available:
https://doi.org/10.3390/horticulturae10111124
[18] A. Bora and P. Sarma, “A Visionary
Analysis on Ways and Techniques Used for Identification of Citrus Plant
Disease,” 2025 3rd International Conference on Intelligent Systems, Advanced
Computing and Communication (ISACC), Silchar, India, Feb. 2025. doi: 10.1109/ISACC65211.2025.10969180.
[19] M. Franco-Hern?ndez, V. Mej?a-Dom?nguez,
and Y. Rodriguez-Gallo, “Advanced Image Processing Methods for Early Detection
and Monitoring of Aphid Infestations,” 2024 IEEE Central America and Panama
Student Conference (CONESCAPAN), Panama, Panama, Sep. 2024. doi:
10.1109/CONESCAPAN62181.2024.10891106.
[20] N. Valarmathi, M. Bharathi Kannan, M.
Elankumaran, M. Gowri Shankar, and B. Gowtham, “Fruit Disease Prediction with
Fertilizer Recommendation for Citrus Family using Deep Learning,” 2023
International Conference on Sustainable Computing and Data Communication
Systems (ICSCDS), Erode, India, Mar. 2023. doi:
10.1109/ICSCDS56580.2023.10104601.
[21] H. Zhu, D. Wang, Y. Wei, P. Wang, and M.
Su, “YOLOV8-CMS: A high-accuracy deep learning model for automated citrus leaf
disease classification and grading,” Plant Methods, vol. 21, no. 1, p. 88,
Jun. 2025, doi: 10.1186/s13007-025-01396-3.
[22] Q. Frederick, T. Burks, J. A. Watson, P.
K. Yadav, J. Qin, M. Kim, and M. M. Dewdney, “Supervised hyperspectral band
selection using texture features for classification of citrus leaf diseases
with YOLOv8,” Sensors, vol. 25, no. 4, p. 1034, Feb. 2025, doi:
10.3390/s25041034.
[23] J. Zhu, J. Chen, H. He, W. Bai, and T.
Zhou, “CBACA-YOLOv5: A symmetric and asymmetric attention-driven detection
framework for citrus leaf disease identification,” Symmetry, vol. 17, no. 4,
p. 617, Apr. 2025, doi: 10.3390/sym17040617.
[24] Q. Frederick, T. Burks, P. K. Yadav, J.
Qin, M. Kim, and M. Dewdney, “Classifying adaxial and abaxial sides of
diseased citrus leaves with selected hyperspectral bands and YOLOv8,” Smart
Agricultural Technology, vol. 9, p. 100600, Dec. 2024, doi: 10.1016/j.atech.2024.100600.
[25] Z. Xiao, C. Yang, Z. Wei, X. Fu, C. Zhao, and G. Li, “A spectral–physiological feature fusion model for the early detection of anthracnose in citrus leaves,” Computers and Electronics in Agriculture, vol. 243, p. 111410, Mar. 2026, doi: 10.1016/j.compag.2026.111410.