Domain-Adaptive Leaf Disease Diagnosis via Customized MobileNetV3-Small: XAI-Guided Severity Scoring and Edge-Deployable API Across Papaya and Lemon

Authors

  • Mohammad Kamrul Hasan Noakhali Science and Technology University image/svg+xml
  • Shourav Dey Noakhali Science and Technology University image/svg+xml
  • Aishwarya Debnath Ayshi Kyoto College of Graduate Studies for Informatics image/svg+xml
  • Md Mahbubul Alam Noakhali Science and Technology University image/svg+xml
  • Ilias Uddin Rifat Noakhali Science and Technology University image/svg+xml
  • Md Ashraful Islam Noakhali Science and Technology University image/svg+xml

DOI:

https://doi.org/10.53799/ncym2e94

Keywords:

Crop disease detection, Transfer Learning, Customized MobileNetV3-Small, Explainable AI (XAI), Edge deployment, Precision agriculture

Abstract

Timely and accurate identification of crop leaf diseases is critical for improving agricultural productivity in developing regions such as Bangladesh, where smallholder farmers suffer yield losses due to delayed or inaccessible diagnosis. This study proposes a transfer learning-based convolutional neural network (CNN) framework for automated papaya leaf disease detection, evaluating Customized MobileNetV3-Small, ResNet50V2, and DenseNet201, fine-tuned on 10,677 papaya leaf images, achieving accuracies of 98.61%, 98.51%, and 99.69%, respectively. Despite DenseNet201 achieving the highest accuracy, its 385 MB model size renders it infeasible for edge deployment on resource-constrained farmer devices. To assess cross-crop generalizability, zero-shot method was validated on a lemon leaf dataset, attaining 97.91% accuracy. Customized MobileNetV3-Small was selected for deployment owing to its compact model size (3.24 MB vs. 6.07 MB standard baseline), low inference latency, and suitability for resource-constrained environments. The model was converted to TFLite format for edge compatibility, with prediction interpretability addressed through LIME-based Explainable AI (XAI) and severity scoring. Integrated into a Flask API prototype, the framework enables real-time, in-field disease diagnosis and supports practical, sustainable crop management effectively in underserved agricultural communities.

Author Biographies

  • Mohammad Kamrul Hasan, Noakhali Science and Technology University

    Mohammad Kamrul Hasan is working as an Assistant Professor of Information and Communication Technology (ICE) at Noakhali Science and Technology University (NSTU) since 24th January 2022. Previously he worked as a Lecturer of IT at University of Information Technology and Sciences (UITS) from 23rd May 2021 to 23rd January 2022. Additionally, he is playing the role as an Assistant Provost of Hazrat Bibi Khadiza Hall of NSTU. He is involved in some research work in the field of Machine Learning, Ad-Hoc Network, Computer Vision, Cloud Computing, and he has published some work based on these topics. Mr. Kamrul completed his BSc.(Engg.) and MSc.(Engg.) degree in Information and Communication Technology from Comilla University, Cumilla-3506, Bangladesh and has achieved a CGPA of 3.96 in both degrees. He has achieved the first position for 7 semesters out 8 in his Bachelor studies. He was awarded UGC merit scholarship in his master's studies. For the outstanding result, Mr. Kamrul has been awarded Chancellor Gold Medal and Prime Minister Gold Medal. Also he has been in the Dean's List award for two times.

  • Shourav Dey, Noakhali Science and Technology University

    Shourav Dey received his B.Sc. (Engg.) degree in Information and Communication Engineering from Noakhali Science and Technology University (NSTU), Bangladesh, in 2025, where he currently works as a research collaborator. He is also an Assistant Teacher of ICT at Bhulta Uchyamadhyamik Bidyalay. His research interests include Artificial Intelligence, Deep Learning, and Computer Vision, focusing on explainable deep ensemble models for smart agriculture and healthcare. He has published first-author research in Elsevier's Smart Agricultural Technology. Additionally, he is the founder and COO of Code For All (CFA), a voluntary organization teaching programming to rural children.

  • Aishwarya Debnath Ayshi, Kyoto College of Graduate Studies for Informatics

    Aishwarya Debnath Ayshi received her B.Sc. degree in Software Engineering from American International University-Bangladesh (AIUB) and her M.Sc. degree in Information Technology from The Kyoto College of Graduate Studies for Informatics, Japan. She is currently a research student engaged in applied machine learning studies. Her research interests include artificial intelligence, machine learning, and computer vision, with applications in dry fish classification and cattle breed recognition, along with statistical and machine learning-based data analysis in her thesis work. She has experience in developing deep learning-based classification systems using transfer learning. She has published peer-reviewed research in Springer Nature–indexed conference proceedings and is currently working on journal submissions.

  • Md Mahbubul Alam, Noakhali Science and Technology University

    Md. Mahbubul Alam received the B.Sc. Eng. and M.Sc. Eng. degrees in Information and Communication Engineering (ICE) from Noakhali Science and Technology University, Bangladesh, in 2016 and 2018, respectively. He has been serving as an Assistant Professor in the Department of Information and Communication Engineering at Noakhali Science and Technology University since 2019 and has over seven years of teaching experience. Prior to this position, he served as a Lecturer in the Department of Computer Science and Engineering at Feni University. He is actively engaged in teaching, research, and academic activities in the field of Information and Communication Engineering. His research interests include machine learning, deep learning, image processing, and data mining. His current research focuses on developing intelligent data-driven models for pattern recognition, computer vision, medical image analysis, and predictive analytics for real-world applications.

  • Ilias Uddin Rifat, Noakhali Science and Technology University

    Ilias Uddin Rifat received his B.Sc. (Eng.) degree in Information and Communication Engineering from Noakhali Science and Technology University (NSTU), Bangladesh. At the undergraduate level, he has been involved in the research work of an explainable ensemble machine learning approach for cardiovascular disease prediction from clinical data from Bangladesh. Currently he is an Assistant Teacher (ICT) at Protiva Public School. He has done industrial training at Bangladesh Telecommunications Company Limited (BTCL) and Professional Training on Data Analysis using Python from the EDGE Program. His research interests include machine learning, explainable AI, data analytics, healthcare informatics, networking and information and communication technologies. Currently, his research interests are on interpretable machine learning models and data-driven solutions for healthcare applications.

  • Md Ashraful Islam, Noakhali Science and Technology University

    Md. Ashraful Islam is currently pursuing the B.Sc. (Eng.) degree in Information and Communication Engineering at Noakhali Science and Technology University, Bangladesh. His research interests include network security, intrusion detection systems, machine learning, and large language model (LLM) applications in cybersecurity. He is actively involved in developing intelligent Network Intrusion Detection Systems (NIDS) that integrate ensemble machine learning models with LLM-based log analysis for enhanced threat detection and classification. His current work focuses on building a unified intrusion detection framework using benchmark datasets.He has also conducted research on feature engineering, data leakage prevention in machine learning-based security systems, cross-dataset model generalization, and real-time network traffic analysis using tools such as Suricata, CICFlowMeter, and Wireshark.

References

[1] B. Koul, B. Pudhuvai, C. Sharma, A. Kumar, V. Sharma, and D. Yadav, “Carica papaya L.: A tropical fruit with benefits beyond the tropics,” Diversity, vol. 14, no. 8, p. 683, 2022, doi: 10.3390/d14080683.

[2] A. F. Bayu, A. Shibiru, and M. A. Aga, “The effect of papaya post-harvest losses on food security of producing households: The case of Dugda Wereda, East Shoa, Ethiopia,” 2025, doi: 10.21203/rs.3.rs-7110032/v1.

[3] M. T. Habib, A. Majumder, A. Z. M. Jakaria, M. Akter, M. S. Uddin, and F. Ahmed, “Machine vision-based papaya disease recognition,” J. King Saud Univ. Comput. Inf. Sci., vol. 32, no. 3, pp. 300–309, 2020, doi: 10.1016/j.jksuci.2018.06.006.

[4] “Classification and identification of leaf disease using digital image processing,” in Chapter 20, 2023, doi: 10.59646/csebookc20/004.

[5] C. N. Sujatha and V. Padmavathi, “Machine learning application for detecting leaf diseases with image processing schemes,” in Machine Learning Methods for Signal, Image and Speech Processing, 2022, pp. 47–76, doi: 10.1201/9781003338789-3.

[6] N. S. Elmitwally, M. Tariq, M. A. Khan, M. Ahmad, S. Abbas, and F. M. Alotaibi, “Rice leaves disease diagnose empowered with transfer learning,” Comput. Syst. Sci. Eng., vol. 42, no. 3, pp. 1001–1014, 2022, doi: 10.32604/csse.2022.022017.

[7] S. Bukhari, “Explainable AI (XAI): Making AI decisions transparent and trustworthy,” 2025, doi: 10.36227/techrxiv.175339404.44374664/v1.

[8] “LEAF applied to evaluate the quality of the explanations provided by LIME and SHAP on the same data point x, using the same ML model f,” doi: 10.7717/peerjcs.479/fig-3.

[9] S. Singh, Z. Furtado, and A. Patil, “AI-driven plant disease detection: Leveraging deep learning for accurate plant disease detection from leaf images,” Int. J. Sci. Res., vol. 13, no. 8, pp. 893–900, 2024, doi: 10.21275/sr24812105523.

[10] “Algorithm 6: Calculating disease severity score,” doi: 10.7717/peerj-cs.1626/table-14.

[11] “Sugarcane leaf disease classification using TensorFlow Lite API,” J. Trends Chall. Artif. Intell., vol. 2, no. 2, 2025, doi: 10.61552/jai.2025.02.002.

[12] D. Faye, I. Diop, N. Mbaye, D. Dione, and M. M. Diedhiou, “Mango fruit diseases severity estimation based on image segmentation and deep learning,” 2024, doi: 10.21203/rs.3.rs-4395003/v1.

[13] S. C. Izah, “Studies on the synergistic effectiveness of methanolic extract of leaves and roots of Carica papaya L. (papaya) against some bacterial pathogens,” Int. J. Complement. Altern. Med., vol. 11, no. 6, 2018, doi: 10.15406/ijcam.2018.11.00429.

[14] K. Tamai, “Effects of environmental factors and soil properties on topographic variations of soil respiration,” Biogeosciences, vol. 7, no. 3, pp. 1133–1142, 2010, doi: 10.5194/bg-7-1133-2010.

[15] P. Maski and A. Thondiyath, “Plant disease detection using advanced deep learning algorithms: A case study of papaya ring spot disease,” in Proc. 6th Int. Conf. Image, Vision and Computing (ICIVC), 2021, doi: 10.1109/ICIVC52351.2021.9526944.

[16] S. Sharmin, M. S. Islam, and M. K. Hasan, “Socioeconomic analysis of alternative farming systems in improving livelihood security of small farmers in selected areas of Bangladesh,” Agriculturists, vol. 10, no. 1, pp. 51–63, 2012, doi: 10.3329/agric.v10i1.11065.

[17] P. A. Chapman, “Reducing product losses in the food supply chain,” in Delivering Performance in Food Supply Chains, 2010, pp. 225–242, doi: 10.1533/9781845697778.3.225.

[18] S. B. Fracker, “Progressive intensification of uncontrolled plant-disease outbreaks,” J. Econ. Entomol., vol. 29, no. 5, pp. 923–940, 1936, doi: 10.1093/jee/29.5.923.

[19] “Transfer learning theory,” in Transfer Learning, 2020, pp. 141–150, doi: 10.1017/9781139061773.012.

[20] “Model-based transfer learning,” in Transfer Learning, 2020, pp. 45–57, doi: 10.1017/9781139061773.006.

[21] A. Dera, “Lightweight neural networks for context aware embedded system,” 2021, doi: 10.36227/techrxiv.16743901.

[22] C. M. Parab, “Lightweight AI models for real-time object detection on edge devices,” 2025, doi: 10.2139/ssrn.5424294.

[23] P. Kulkarni and S. Shastri, “Rice leaf diseases detection using machine learning,” J. Sci. Res. Technol., pp. 17–22, 2024, doi: 10.61808/jsrt81.

[24] A. S. Kini, K. V. Prema, and S. N. Pai, “Early stage black pepper leaf disease prediction based on transfer learning using ConvNets,” Sci. Rep., vol. 14, no. 1, 2024, doi: 10.1038/s41598-024-51884-0.

[25] “Potato leaf disease identification using machine learning,” Int. J. Sci. Res. Eng. Manag., vol. 7, no. 11, pp. 1–11, 2023, doi: 10.55041/IJSREM27322.

[26] S. Sagar and J. Singh, “An experimental study of tomato viral leaf diseases detection using machine learning classification techniques,” Bull. Electr. Eng. Inform., vol. 12, no. 1, pp. 451–461, 2023, doi: 10.11591/eei.v12i1.4385.

[27] S. E. Arman et al., “BananaLSD: A banana leaf images dataset for classification of banana leaf diseases using machine learning,” Data Brief, vol. 50, p. 109608, 2023, doi: 10.1016/j.dib.2023.109608.

[28] K. Kethineni and G. Pradeepini, “Identification of leaf disease using machine learning algorithm for improving the agricultural system,” J. Adv. Inf. Technol., vol. 14, no. 1, pp. 122–129, 2023, doi: 10.12720/jait.14.1.122-129.

[29] K. Prabavathy et al., “Plant leaf disease detection using machine learning,” in Proc. 2nd Int. Conf. Applied Artificial Intelligence and Computing (ICAAIC), 2023, pp. 378–382, doi: 10.1109/ICAAIC56838.2023.10140367.

[30] A. Mimi et al., “Identifying selected diseases of leaves using deep learning and transfer learning models,” Mach. Graph. Vis., vol. 32, no. 1, pp. 55–71, 2023, doi: 10.22630/mgv.2023.32.1.3.

[31] A. Saeed et al., “Smart detection of tomato leaf diseases using transfer learning-based convolutional neural networks,” Agriculture, vol. 13, no. 1, p. 139, 2023, doi: 10.3390/agriculture13010139.

[32] O. Attallah, “Tomato leaf disease classification via compact convolutional neural networks with transfer learning and feature selection,” Horticulturae, vol. 9, no. 2, p. 149, 2023, doi: 10.3390/horticulturae9020149.

[33] W. Wu, “Identification of tea leaf diseases based on deep transfer learning,” Front. Comput. Intell. Syst., vol. 2, no. 3, pp. 75–77, 2023, doi: 10.54097/fcis.v2i3.5218.

[34] R. Zhang, Y. Wang, P. Jiang, J. Peng, and H. Chen, “IBSA_Net: A network for tomato leaf disease identification based on transfer learning with small samples,” Appl. Sci., vol. 13, no. 7, p. 4348, 2023, doi: 10.3390/app13074348.

[35] A. S. Hatem, M. S. Altememe, and M. A. Fadhel, “Identifying corn leaves diseases by extensive use of transfer learning: A comparative study,” Indones. J. Electr. Eng. Comput. Sci., vol. 29, no. 2, p. 1030, 2023, doi: 10.11591/ijeecs.v29.i2.pp1030-1038.

[36] R. S. R. Singh and R. K. Sanodiya, “Zero-shot transfer learning framework for plant leaf disease classification,” IEEE Access, vol. 11, pp. 143861–143880, 2023, doi: 10.1109/ACCESS.2023.3343759.

[37] A. Emmanuel et al., “Classification of cassava leaf diseases using deep Gaussian transfer learning model,” 2022, doi: 10.22541/au.166420984.44354671/v1.

[38] M. S. A. M. Al-gaashani et al., “Tomato leaf disease classification by exploiting transfer learning and feature concatenation,” IET Image Process., vol. 16, no. 3, pp. 913–925, 2022, doi: 10.1049/ipr2.12397.

[39] X. Yu, Q. Gong, C. Chen, and L. Lu, “Automatic diagnosis of soybean leaf disease by transfer learning,” Am. J. Biochem. Biotechnol., vol. 18, no. 2, pp. 252–260, 2022, doi: 10.3844/ajbbsp.2022.252.260.

[40] A. Nader, M. H. Khafagy, and S. A. Hussien, “Grape leaves diseases classification using ensemble learning and transfer learning,” Int. J. Adv. Comput. Sci. Appl., vol. 13, no. 7, 2022, doi: 10.14569/IJACSA.2022.0130767.

[41] V. Gautam et al., “A transfer learning-based artificial intelligence model for leaf disease assessment,” Sustainability, vol. 14, no. 20, p. 13610, 2022, doi: 10.3390/su142013610.

[42] M. Fraiwan, E. Faouri, and N. Khasawneh, “Classification of corn diseases from leaf images using deep transfer learning,” Plants, vol. 11, no. 20, p. 2668, 2022, doi: 10.3390/plants11202668.

[43] P. Pradhan, B. Kumar, and S. Mohan, “Comparison of various deep convolutional neural network models to discriminate apple leaf diseases using transfer learning,” J. Plant Dis. Prot., vol. 129, no. 6, pp. 1461–1473, 2022, doi: 10.1007/s41348-022-00660-1.

[44] M. G. Yadav, R. Nennuri, D. Rajeshwari, V. Rishitha, and T. Puneeth, “Identification of plant leaf diseases using machine learning algorithms,” Ann. Rom. Soc. Cell Biol., vol. 25, no. 6, pp. 6866–6875, 2021.

[45] Z. Jiang, Z. Dong, W. Jiang, and Y. Yang, “Recognition of rice leaf diseases and wheat leaf diseases based on multi-task deep transfer learning,” Comput. Electron. Agric., vol. 186, p. 106184, 2021, doi: 10.1016/j.compag.2021.106184.

[46] K. N et al., “Rice leaf diseases prediction using deep neural networks with transfer learning,” Environ. Res., vol. 198, p. 111275, 2021, doi: 10.1016/j.envres.2021.111275.

[47] S. Dey, A. D. Ayshi, and A. R. Chowdhury, “AI-driven framework for cattle breed identification to support sustainable livestock and One Health,” in Advances in Social Science, Education and Humanities Research, 2026, pp. 258–269, doi: 10.2991/978-2-38476-581-2_20.

[48] U. S. Rao et al., “Deep learning precision farming: Grapes and mango leaf disease detection by transfer learning,” Glob. Transit. Proc., vol. 2, no. 2, pp. 535–544, 2021, doi: 10.1016/j.gltp.2021.08.002.

[49] M. Hasan, B. Tanawala, and K. J. Patel, “Deep learning precision farming: Tomato leaf disease detection by transfer learning,” SSRN Electron. J., 2019, doi: 10.2139/ssrn.3349597.

[50] S. Mustofa, M. T. Ahad, Y. R. Emon, and A. Sarker, “BDPapayaLeaf: A dataset of papaya leaf for disease detection, classification, and analysis,” Data Brief, vol. 57, p. 110910, 2024, doi: 10.1016/j.dib.2024.110910.

[51] R. Gani et al., “Smartphone image dataset to distinguish healthy and unhealthy leaves in papaya orchards in Bangladesh,” Data Brief, vol. 55, p. 110599, 2024, doi: 10.1016/j.dib.2024.110599.

[52] M. Shaheen, “Lemon Leaf Disease Dataset (LLDD),” Kaggle, 2023. [Online]. Available: https://www.kaggle.com/datasets/mahmoudshaheen1134/lemon-leaf-disease-dataset-lldd (accessed Nov. 16, 2024).

[53] A. Upreti, “Convolutional neural network (CNN). A comprehensive overview,” 2022, doi: 10.20944/preprints202208.0313.v1.

[54] M. Basireddy and R. Dayana, “Leaf disease identification with MobileNetV3 small: A lightweight approach,” in Proc. Int. Conf. Communication, Computing, Smart Materials and Devices (ICCCSMD), 2024, pp. 1–8, doi: 10.1109/ICCCSMD63546.2024.11015227.

[55] N. R. Raje and A. Jadhav, “Automated diagnosis of pneumonia through capsule network in conjunction with ResNet50v2 model,” in Proc. Int. Conf. Emerging Smart Computing and Informatics (ESCI), 2022, pp. 1–6, doi: 10.1109/ESCI53509.2022.9758184.

[56] H. Dong, L. Zhang, and B. Zou, “Densely connected convolutional neural network based polarimetric SAR image classification,” in Proc. IEEE Int. Geoscience and Remote Sensing Symp. (IGARSS), 2019, doi: 10.1109/IGARSS.2019.8900292.

[57] S. Dey, A. D. Ayshi, and A. R. Chowdhury, “AI-driven framework for cattle breed identification to support sustainable livestock and One Health,” in Advances in Social Science, Education and Humanities Research, 2026, pp. 258–269, doi: 10.2991/978-2-38476-581-2_20.

[58] J. A. Bacus and N. B. Linsangan, “Detection and identification with analysis of Carica papaya leaf using Android,” J. Adv. Inf. Technol., vol. 13, no. 2, pp. 162–166, 2022, doi: 10.12720/jait.13.2.162-166.

[59] A. S. Chaithanya and M. Rachana, “Identification of diseased papaya leaf through transfer learning,” Indian J. Sci. Technol., vol. 16, no. 48, pp. 4676–4687, 2023, doi: 10.17485/ijst/v16i48.2690.

[60] S. Dey, M. K. Hasan, A. Adhikary, S. Akter, and M. S. Ejaz, “SwinGhost-ClustNet: An explainable deep ensemble model for papaya leaf disease detection and field deployment in Bangladeshi agriculture,” Smart Agric. Technol., vol. 9, p. 101824, 2026, doi: 10.1016/j.atech.2026.101824.

[61] J. G. Horsfall and R. W. Barratt, “An improved grading system for measuring plant diseases,” Phytopathology, vol. 35, no. 5, p. 655, 1945.

[62] M. Shaheen, M. Bashir, A. Zafar, and M. Arshad, “Lightweight deep learning approach for early detection of lemon leaf diseases using enhanced MobileNetV2,” J. Inf. Syst. Eng. Manag., vol. 10, no. 3, p. 12791, 2025, doi: 10.55267/jisem.12791.

[63] D. Sutaji and U. Yusoff, “LEMOXINET: Lite ensemble MobileNetV2 and Xception architecture for plant disease classification,” Comput. Electron. Agric., vol. 202, p. 107398, 2022, doi: 10.1016/j.compag.2022.107398.

[64] M. R. Islam, M. A. Rahman, and S. Ahmed, “BDLemonLeaf: Image dataset of healthy and disease lemon (Citrus limon) leaves from Bangladesh,” Mendeley Data, 2024, doi: 10.17632/643f5bbc2t.1.

[65] M. A. Rahman, S. Islam, and T. Ahmed, “Lemon leaf dataset: Images of Citrus diseases from Dhaka, Bangladesh,” Mendeley Data, 2025, doi: 10.17632/smz8ffbthy.3.

[66] S. Karim, M. Hussain, A. Khan, and M. Iqbal, “Citrus diseases detection using innovative deep learning approach with DenseNet-201 and AlexNet,” PLoS ONE, vol. 20, no. 1, p. e0316081, 2025, doi: 10.1371/journal.pone.0316081.

Downloads

Published

31-07-2026

Issue

Vol. 24 No. 3 (2026): AJSE Volume 24 Issue 3

Section

Articles

License

Copyright (c) 2026 AIUB Journal of Science and Engineering

Creative Commons License

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

AJSE contents are under the terms of the Creative Commons Attribution License. This permits anyone to copy, distribute, transmit and adapt the work non-commercially provided the original work and source is appropriately cited.

How to Cite

[1]
“Domain-Adaptive Leaf Disease Diagnosis via Customized MobileNetV3-Small: XAI-Guided Severity Scoring and Edge-Deployable API Across Papaya and Lemon”, AJSE, vol. 24, no. 3, pp. 307–317, Jul. 2026, doi: 10.53799/ncym2e94.