Enhanced Plant Leaf Disease Detection Using CLAHE-Processed Images and Custom CNN Deep Learning Architecture

Abstract

This study aims to develop an efficient and accurate deep learning-based model for the classification of plant leaf diseases using Convolutional Neural Networks (CNN). The objective is to automate disease detection in agricultural crops to assist farmers and agricultural experts in early and reliable diagnosis. The model is trained on the publicly available “Plant Village CLAHE Processed Data” dataset, which includes high-resolution RGB images of healthy and diseased plant leaves. Images are preprocessed through resizing (128×128), normalized, and split into training, validation, and test sets. Data augmentation techniques such as flipping, zooming, and rotation are used to improve generalization. A custom CNN architecture comprising convolutional, pooling, dense, and dropout layers is employed and trained using the Adam optimizer. Exploratory Data Analysis (EDA) ensures data quality and balance. The model achieves impressive results, with 93% test accuracy, 91% precision, 93% recall, and an F1-score of 92%, indicating robust performance in identifying diverse plant diseases. Training accuracy reached 94.64% with a validation accuracy of 92.95%, confirming minimal overfitting. These results validate the model’s reliability for practical use in smart farming solutions, especially in mobile or IoT-based applications for real-time disease monitoring and precision agriculture.

Introduction

Agriculture is crucial for global food security and economies, especially in farming-dependent countries. However, plant diseases cause significant crop losses—estimated by the FAO at 20–40% annually—impacting yield, quality, and costs. Early and accurate disease detection is essential to mitigate these effects. Traditional manual inspection methods are laborious, subjective, and inaccessible in many remote areas.

Deep learning, particularly Convolutional Neural Networks (CNNs), has revolutionized image-based plant disease diagnosis by automatically learning features from leaf images. CNNs inspired by human vision efficiently detect subtle disease symptoms like discoloration and texture changes, even in early stages. Popular CNN architectures (e.g., LeNet, AlexNet, VGGNet, ResNet) have been successfully applied to various crops including tomatoes, potatoes, maize, grapes, apples, and rice, often achieving classification accuracies above 90%.

These CNN models are integrated into practical tools such as mobile apps, drones, and IoT devices, enabling farmers—especially smallholders—to access rapid, scalable disease diagnosis and surveillance. This democratizes agricultural knowledge and facilitates proactive management. However, challenges remain including the need for diverse, high-quality training data that reflect real-field conditions, model interpretability for end users, and infrastructural barriers in rural areas.

Explainable AI techniques are being developed to enhance model transparency and user trust. Successful deployment requires collaboration among researchers, farmers, governments, and industry, supported by policies promoting data access, technology subsidies, and training. CNN-based plant disease detection holds transformative potential to advance precision agriculture and global food sustainability.

The literature review highlights recent advances such as YOLOv8n for tomato disease detection, hybrid CNN models with optimization techniques, and explainable AI approaches that balance accuracy with interpretability. Various studies report accuracies ranging from 95% to over 99%, demonstrating the practical feasibility of deep learning in agriculture.

The described methodology involves using a CLAHE-processed PlantVillage dataset, with image preprocessing, augmentation, and a custom CNN architecture trained using the Adam optimizer. Performance is evaluated using accuracy, precision, recall, and F1-score, emphasizing balanced and efficient model training.

Conclusion

This work uses a proprietary convolutional neural network (CNN) to show an efficient and accurate deep learning method for the classification of plant leaf diseases. Using the PlantVillage dataset processed by CLAHE, the model solves typical constraints in agricultural disease diagnosis like low picture contrast and model overfitting. To improve generalisation, the approach consists in important processes including image scaling, normalisation, and data augmentation methods including flipping, zooming, and rotation. Trained using the Adam optimiser, a well crafted CNN architecture including convolutional, pooling, dense, and dropout layers guarantees effective feature extraction and model stability. With precision of 91%, recall of 93%, and an F1-score of 92%, the model attained a test accuracy of 93% suggesting strong and balanced performance. With low loss values, training and validation accuracies of 94.64% and 92.95% respectively demonstrate minimum overfitting and dependable learning.

References

[1] J. Liu and X. Wang, “Plant diseases and pests detection based on deep learning?: a review,” Plant Methods, pp. 1–18, 2021, doi: 10.1186/s13007-021-00722-9. [2] Y. Guo et al., “Plant Disease Identification Based on Deep Learning Algorithm in Smart Farming,” Discret. Dyn. Nat. Soc., vol. 2020, 2020, doi: 10.1155/2020/2479172. [3] K. Golhani, S. K. Balasundram, G. Vadamalai, and B. Pradhan, “A review of neural networks in plant disease detection using hyperspectral data,” Inf. Process. Agric., vol. 5, no. 3, pp. 354–371, 2018, doi: 10.1016/j.inpa.2018.05.002. [4] K. Zhang, Q. Wu, A. Liu, and X. Meng, “Can deep learning identify tomato leaf disease?,” Adv. Multimed., vol. 2018, 2018, doi: 10.1155/2018/6710865. [5] S. Prajapati, S. Qureshi, Y. Rao, S. Nadkarni, M. Retharekar, and A. Avhad, “Plant Disease Identification Using Deep Learning,” 2023 4th Int. Conf. Emerg. Technol. INCET 2023, vol. 90, no. February, pp. 249–257, 2023, doi: 10.1109/INCET57972.2023.10170463. [6] Y. M. Abd Algani, O. J. Marquez Caro, L. M. Robladillo Bravo, C. Kaur, M. S. Al Ansari, and B. Kiran Bala, “Leaf disease identification and classification using optimized deep learning,” Meas. Sensors, vol. 25, no. September 2022, p. 100643, 2023, doi: 10.1016/j.measen.2022.100643. [7] S. S. Bhoomika and K. M. Poornima, “Plant Leaf Disease Detection and Classification Using Deep Learning Technique,” Lect. Notes Networks Syst., vol. 494, no. 3, pp. 73–83, 2023, doi: 10.1007/978-981-19-4863-3_7. [8] M. Bouni, B. Hssina, K. Douzi, and S. Douzi, “Impact of Pretrained Deep Neural Networks for Tomato Leaf Disease Prediction,” J. Electr. Comput. Eng., vol. 2023, 2023, doi: 10.1155/2023/5051005. [9] B. N. Kahsay and O. D. Makinde, “Ecoepidemiological Model and Optimal Control Analysis of Tomato Yellow Leaf Curl Virus Disease in Tomato Plant,” J. Appl. Math., vol. 2023, 2023, doi: 10.1155/2023/4066236. [10] K. Academy, E. Balakannan, K. Academy, and E. Maragatharajan, “A Novel Method for Predicting Plant Leaf Disease Based on Machine Learning and Deep Learning Techniques,” no. Ml, pp. 1–24, 2023. [11] V. Balaji et al., “Deep Transfer Learning Technique for Multimodal Disease Classification in Plant Images,” Contrast Media Mol. Imaging, vol. 2023, p. 5644727, 2023, doi: 10.1155/2023/5644727. [12] B. Panigrahi, K. C. R. Kathala, and M. Sujatha, “A Machine Learning-Based Comparative Approach to Predict the Crop Yield Using Supervised Learning With Regression Models,” Procedia Comput. Sci., vol. 218, no. 2022, pp. 2684–2693, 2023, doi: 10.1016/j.procs.2023.01.241. [13] D. Paudel et al., “Machine learning for regional crop yield forecasting in Europe,” F. Crop. Res., vol. 276, no. December 2021, p. 108377, 2022, doi: 10.1016/j.fcr.2021.108377. [14] A. Tripathi, R. K. Tiwari, and S. P. Tiwari, “A deep learning multi-layer perceptron and remote sensing approach for soil health based crop yield estimation,” Int. J. Appl. Earth Obs. Geoinf., vol. 113, no. April, p. 102959, 2022, doi: 10.1016/j.jag.2022.102959. [15] K. Jhajharia, P. Mathur, S. Jain, and S. Nijhawan, “Crop Yield Prediction using Machine Learning and Deep Learning Techniques,” Procedia Comput. Sci., vol. 218, pp. 406–417, 2022, doi: 10.1016/j.procs.2023.01.023. [16] K. Joshi et al., “Precision diagnosis of tomato diseases for sustainable agriculture through deep learning approach with hybrid data augmentation,” Curr. Plant Biol., vol. 41, no. August 2024, p. 100437, 2025, doi: 10.1016/j.cpb.2025.100437. [17] S. Pradeep and J. Joseph, “Enhancing pepper growth and yield through disease identification in plants using leaf-based deep learning techniques,” pp. 54–59, 2025, doi: 10.1201/9781003559115-10. [18] H. Habaragamuwa, Y. Oishi, and K. Tanaka, “Achieving explainability for plant disease classification with disentangled variational autoencoders,” Eng. Appl. Artif. Intell., vol. 133, no. PA, p. 107982, 2024, doi: 10.1016/j.engappai.2024.107982. [19] H. H. Alshammari and A. Alzahrani, “Employing a hybrid lion-firefly algorithm for recognition and classification of olive leaf disease in Saudi Arabia,” Alexandria Eng. J., vol. 84, no. October, pp. 215–226, 2023, doi: 10.1016/j.aej.2023.10.057. [20] M. A. Bhatti et al., “Advanced Plant Disease Segmentation in Precision Agriculture using Optimal Dimensionality Reduction with Fuzzy C-Means Clustering and Deep Learning,” IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens., vol. 17, pp. 18264–18277, 2024, doi: 10.1109/JSTARS.2024.3437469. [21] W. Li, L. Zhu, and J. Liu, “PL-DINO: An Improved Transformer-Based Method for Plant Leaf Disease Detection,” Agric., vol. 14, no. 5, pp. 1–14, 2024, doi: 10.3390/agriculture14050691. [22] A. Sharma et al., “Rose Plant Disease Detection Using Image Processing and Machine Learning,” Commun. Comput. Inf. Sci., vol. 2050 CCIS, pp. 69–85, 2024, doi: 10.1007/978-3-031-58953-9_6. [23] D. Mane, M. Deore, R. Ashtagi, S. Shinde, and Y. Gurav, “Basil plant leaf disease detection using amalgam based deep learning models,” J. Auton. Intell., vol. 7, no. 1, pp. 1–13, 2024, doi: 10.32629/jai.v7i1.1002. [24] A. Ouamane et al., “Enhancing Plant Disease Detection: A Novel CNN-Based Approach with Tensor Subspace Learning and HOWSVD-MD,” pp. 1–17, 2024, [Online]. Available: http://arxiv.org/abs/2405.20058 [25] A. Gautam, K. Kinjalk, A. Kumar, and V. Priye, “FBG-Based Respiration Rate Sensing with Arduino Interface,” IEEE Sens. J., vol. 20, no. 16, pp. 9209–9217, Aug. 2020, doi: 10.1109/JSEN.2020.2989004. [26] “(PDF) Plant species recognition using modified LeNet-5 CNN architecture.” https://www.researchgate.net/publication/343471599_Plant_species_recognition_using_modified_LeNet-5_CNN_architecture (accessed Jun. 14, 2025).

Copyright

Copyright © 2025 Neelam Sulaiya, Shyamol Banerjee. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.