A Hybrid Lightweight Convolutional Neural Network Model for Image-Based Robusta Coffee Leaf Diseases Classification

Purwohudiono; Subiyanto; Nektar  Cahayasabda; Galang Putra  Prasetya; Dhitsa Anggara Ari  Samasta

doi:10.15408/jti.v19i2.50557

Authors

Purwohudiono Departement of Electrical Engineering, Faculty of Engineering, Universitas Negeri Semarang, Jawa Tengah, Indonesia
Subiyanto Departement of Electrical Engineering, Faculty of Engineering, Universitas Negeri Semarang, Jawa Tengah, Indonesia https://orcid.org/0000-0001-5890-7017
Nektar Cahayasabda Departement of Electrical Engineering, Faculty of Engineering, Universitas Negeri Semarang, Jawa Tengah, Indonesia
Galang Putra Prasetya Departement of Electrical Engineering, Faculty of Engineering, Universitas Negeri Semarang, Jawa Tengah, Indonesia
Dhitsa Anggara Ari Samasta Departement of Electrical Engineering, Faculty of Engineering, Universitas Negeri Semarang, Jawa Tengah, Indonesia

DOI:

https://doi.org/10.15408/jti.v19i2.50557

Keywords:

Coffe leaf disease, Deep learning, Image classification, Lightweight model

Abstract

Convolutional Neural Networks (CNNs) have been widely used for the classification of coffee leaf diseases. However, the resource-intensive nature of conventional CNNs limits their feasibility for coffee leaf disease identification on edge devices in farming scenarios. To achieve a lightweight, high-accuracy CNN model, this paper develops an hybrid MobileNetV2-NASNetMobile for classifying Robusta coffee leaf diseases. The model integrates two lightweight CNNs: MobileNetV2, which employs depthwise separable convolutions for efficiency, and NASNetMobile, which optimizes discriminative feature extraction. Both branches receive input images resized to 224 × 224 pixels. A feature-level integration via concatenation is then used to combine their independent feature representations, thereby enhancing model robustness and generalization. The training and evaluation of the implemented method were conducted using a dataset of 3,678 Robusta coffee leaf images, sourced from publicly available datasets and field photographs. Model performance was assessed using 5-fold cross-validation to obtain a reliable estimate of generalization and to mitigate bias from a single train-test split. The implemented method attained an accuracy of 99.11%, with corresponding precision of 99.05%, recall of 98.97%, and F1-score of 99.00%. Furthermore, expert validation yielded 85% agreement, a discrepancy from the 99.11% test accuracy attributable to curated test conditions versus real-world variability in lighting, symptom severity, and diagnostic subjectivity. These results nonetheless demonstrate the model’s effectiveness as a scalable solution for Robusta coffee leaf disease classification on resource-limited devices.

References

[1] B. Tugrul, E. Elfatimi, and R. Eryigit, “Convolutional Neural Networks in Detection of Plant Leaf Diseases: A Review,” Aug. 01, 2022, MDPI. doi: 10.3390/agriculture12081192.

[2] T. D. Salka, M. B. Hanafi, S. M. S. A. A. Rahman, D. B. M. Zulperi, and Z. Omar, “Plant leaf disease detection and classification using convolution neural networks model: a review,” Artif. Intell. Rev., vol. 58, no. 10, Oct. 2025, doi: 10.1007/s10462-025-11234-6.

[3] W. Ode Asriyani et al., “Expert System for Coffee Leaf Disease Classification With Convolutional Neural Networks Sistem Pakar Klasifikasi Penyakit Pada Daun Kopi Dengan Metode CNN,” Jurnal Informatika dan Teknologi Informasi, vol. 22, no. 3, pp. 1–12, 2025, doi: 10.31515/telematika.v22i3.14323.

[4] O. Adelaja and B. Pranggono, “Leveraging Deep Learning for Real-Time Coffee Leaf Disease Identification,” AgriEngineering, vol. 7, no. 13, 2025, doi: 10.3390/agriengineering7010013.

[5] I. V. C. Motta, N. Vuillerme, H. H. Pham, and F. A. P. de Figueiredo, “Machine learning techniques for coffee classification: a comprehensive review of scientific research,” Artif. Intell. Rev., vol. 58, no. 15, 2024, doi: 10.1007/s10462-024-11004-w.

[6] C. Calderón-Mosilot, U. Tapia-Gálvez, J. Arcila-Diaz, and H. I. Mejia-Cabrera, “Diagnosis of Nutritional Deficiencies in Coffee Plants Through Automated Analysis of Digital Images Using Deep Learning in Uncontrolled Agricultural Environments,” AgriEngineering, vol. 7, no. 421, 2025, doi: 10.3390/agriengineering7120421.

[7] M. D. Islam et al., “Towards real-time weed detection and segmentation with lightweight CNN models on edge devices,” Comput. Electron. Agric., vol. 237, Jun. 2025, doi: 10.1016/j.compag.2025.110600.

[8] S. Khan, F. Siddiqui, and M. A. Ahad, “Bridging efficiency and interpretability: Explainable AI for multi-classification of pulmonary diseases utilizing modified lightweight CNNs,” Image Vis. Comput., vol. 158, May 2025, doi: 10.1016/j.imavis.2025.105553.

[9] I Gede Rian Lanang Oka, Anak Agung Gede Bagus Ariana, Wayan Sauri Peradhayana, Ni Luh Wiwik Sri Rahayu Ginantra, and I Ketut Sutarwiyasa, “Performance Comparison of MobileNetV2 and NASNetMobile Architectures in Soybean Leaf Disease Classification,” Indonesian Journal of Data and Science, vol. 6, no. 2, pp. 251–259, Jul. 2025, doi: 10.56705/ijodas.v6i2.243.

[10] S. Kulkarni et al., “Coffee Plant Disease Identification using Enhanced Short Learning EfficientNetV2,” in 2023 IEEE 20th India Council International Conference, INDICON 2023, Institute of Electrical and Electronics Engineers Inc., 2023, pp. 91–96. doi: 10.1109/INDICON59947.2023.10440883.

[11] W. Shafik, A. Tufail, C. De Silva Liyanage, and R. A. A. H. M. Apong, “Using transfer learning-based plant disease classification and detection for sustainable agriculture,” BMC Plant Biol., vol. 24, no. 1, Dec. 2024, doi: 10.1186/s12870-024-04825-y.

[12] K. Dhakyanaikabdel and A. J, “A Hybrid CNN-Vision Transformer Framework for Optimized Leaf Disease Detection Using Feature Fusion and Transfer Learning,” Journal of Robotics and Control (JRC), vol. 6, no. 6, p. 2025, doi: 10.18196/jrc.v6i6.28426.

[13] K. Taji, A. Sohail, T. Shahzad, B. S. Khan, M. A. Khan, and K. Ouahada, “An Ensemble Hybrid Framework: A Comparative Analysis of Metaheuristic Algorithms for Ensemble Hybrid CNN Features for Plants Disease Classification,” IEEE Access, vol. 12, pp. 61886–61906, Apr. 2024, doi: 10.1109/ACCESS.2024.3389648.

[14] A. A. Hailu, B. C. Kassa, E. A. Desta, F. B. Adugna, and A. O. Salau, “Ensemble-based sesame disease detection and classification using deep convolutional neural networks (CNN),” Sci. Rep., vol. 15, no. 1, Dec. 2025, doi: 10.1038/s41598-025-08076-1.

[15] A. K. Trivedi, T. Mahajan, T. Maheshwari, R. Mehta, and S. Tiwari, “Leveraging feature fusion ensemble of VGG16 and ResNet-50 for automated potato leaf abnormality detection in precision agriculture,” Soft comput., vol. 29, no. 4, pp. 2263–2277, Feb. 2025, doi: 10.1007/s00500-025-10523-0.

[16] M. S. A. M. Al-gaashani, F. Shang, M. S. A. Muthanna, M. Khayyat, and A. A. Abd El-Latif, “Tomato leaf disease classification by exploiting transfer learning and feature concatenation,” IET Image Process., vol. 16, no. 3, pp. 913–925, Feb. 2022, doi: 10.1049/ipr2.12397.

[17] S. Muthusamy and S. P. Ramu, “A neural architecture search optimized lightweight attention ensemble model for nutrient deficiency and severity assessment in diverse crop leaves,” Sci. Rep., vol. 15, no. 1, Dec. 2025, doi: 10.1038/s41598-025-20124-4.

[18] M. Emin Sahin, U. Özkaya, C. Arisoy, H. İ. Coşar, and H. Ulutaş, “CucuNetCNNs: Application of novel ensemble deep neural networks for classification of cucumber leaf disease,” Ain Shams Engineering Journal, vol. 16, no. 5, Apr. 2025, doi: 10.1016/j.asej.2025.103380.

[19] Y. Aufar and T. P. Kaloka, “Robusta coffee leaf diseases detection based on MobileNetV2 model,” International Journal of Electrical and Computer Engineering, vol. 12, no. 6, pp. 6675–6683, Dec. 2022, doi: 10.11591/ijece.v12i6.pp6675-6683.

[20] S. Chelangat, R. Anirwoth, K. N. Mayanja, and A. Sserwadda, “A Machine Learning Dataset for Classification of Common Coffee Leaf Diseases in Uganda.,” vol. 1, Feb. 2025, doi: 10.17632/K36WND6KNB.1.

[21] T. B. Nguyen-Tat, T. Q. Hung, P. T. Nam, and V. M. Ngo, “Evaluating pre-processing and deep learning methods in medical imaging: Combined effectiveness across multiple modalities,” Alexandria Engineering Journal, vol. 119, pp. 558–586, Apr. 2025, doi: 10.1016/j.aej.2025.01.090.

[22] R. Zhang, B. Zhou, C. Lu, and M. Ma, “The Performance Research of the Data Augmentation Method for Image Classification,” Math. Probl. Eng., vol. 2022, 2022, doi: 10.1155/2022/2964829.

[23] T. Kumar, R. Brennan, A. Mileo, and M. Bendechache, “Image Data Augmentation Approaches: A Comprehensive Survey and Future Directions,” IEEE Access, vol. 12, pp. 187536–187571, Sep. 2024, doi: 10.1109/ACCESS.2024.3470122.

[24] M. M. Musthafa, I. Manimozhi, T. R. Mahesh, and S. Guluwadi, “Optimizing double-layered convolutional neural networks for efficient lung cancer classification through hyperparameter optimization and advanced image pre-processing techniques,” BMC Med. Inform. Decis. Mak., vol. 24, no. 1, Dec. 2024, doi: 10.1186/s12911-024-02553-9.

[25] I. U. W. Mulyono, E. H. Rachmawanto, C. A. Sari, and M. K. Sarker, “A high accuracy of deep learning based CNN architecture: classic, VGGNet, and RestNet50 for Covid-19 image classification,” Telkomnika (Telecommunication Computing Electronics and Control), vol. 22, no. 5, pp. 1187–1195, Oct. 2024, doi: 10.12928/TELKOMNIKA.v22i5.26017.

[26] D. Rastogi, P. Johri, V. Tiwari, and A. A. Elngar, “Multi-class classification of brain tumour magnetic resonance images using multi-branch network with inception block and five-fold cross validation deep learning framework,” Biomed. Signal Process. Control, vol. 88, p. 105602, Feb. 2024, doi: 10.1016/J.BSPC.2023.105602.

[27] T. M. Lam, S. N. Prova, T. e. Mustafa, and N. A. Khan, “Improving Lung Cancer Prediction Using Advanced Hyperparameter Optimization and Nested Cross-Validation,” Engineering Reports, vol. 7, no. 11, p. e70486, Nov. 2025, doi: 10.1002/ENG2.70486;WGROUP:STRING:PUBLICATION.

[28] C. E. S. E. Silva, J. B. Fragoso, T. Paixao, A. B. Alvarez, and F. Palomino-Quispe, “A Low Computational Cost Deep Learning Approach for Localization and Classification of Diseases and Pests in Coffee Leaves,” IEEE Access, vol. 13, pp. 71943–71964, 2025, doi: 10.1109/ACCESS.2025.3562832.

[29] B. Mehta, S. Bharany, D. H. Elkamchouchi, A. U. Rehman, R. Singh, and S. H. Adem, “EffResViT-SE FusionNet: A Hybrid Deep Learning Framework for Accurate Classification of Coffee Leaf Diseases,” Food Sci. Nutr., vol. 13, no. 12, Dec. 2025, doi: 10.1002/fsn3.71311.

[30] Z. Zhao, E. B. A. Bakar, N. B. A. Razak, and M. N. Akhtar, “Progressive optimization of EfficientNetV2 for classification based on images of corroded objects,” Computational and Applied Mathematics, vol. 44, no. 419, Aug. 2025, doi: 10.1007/s40314-025-03382-w.

[31] J. R. Korsipati, R. M. R. Yanamala, A. Pallakonda, R. D. A. Raj, and K. K. Prakasha, “Multi-resolution transfer learning for tampered image classification using SE-enhanced fused-MBConv and optimized CNN heads,” Sci. Rep., vol. 15, no. 1, Dec. 2025, doi: 10.1038/s41598-025-17799-0.

[32] R. Tobiasz, G. Wilczyński, P. Graszka, N. Czechowski, and S. Łuczak, “Edge Devices Inference Performance Comparison,” Jun. 2023, doi: 10.5626/JCSE.2023.17.2.51.

[33] M. Reda, R. Suwwan, S. Alkafri, Y. Rashed, and T. Shanableh, “AgroAId: A Mobile App System for Visual Classification of Plant Species and Diseases Using Deep Learning and TensorFlow Lite,” Informatics, vol. 9, no. 3, Sep. 2022, doi: 10.3390/informatics9030055.

[34] M. Sandler, A. Howard, M. Zhu, A. Zhmoginov, and L.-C. Chen, “MobileNetV2: Inverted Residuals and Linear Bottlenecks,” Mar. 2019, [Online]. Available: http://arxiv.org/abs/1801.04381

[35] B. Zoph, V. Vasudevan, J. Shlens, and Q. V. Le, “Learning Transferable Architectures for Scalable Image Recognition,” Apr. 2018, [Online]. Available: http://arxiv.org/abs/1707.07012

[36] W. M. Alawad, A. Alharbi, S. Alanazi, T. Almotairy, A. I. Alhujaylan, and K. E. Hamed, “Development of a Certainty Factor–Based Mobile Expert System for Palm Tree Pest and Disease Classification,” Univers. J. Agric. Res., vol. 13, no. 5, pp. 68–80, 2025, doi: 10.13189/ujar.2025.130502.