Rice Leaf Diseases Classification Using CNN with Transfer Learning

Abstract

Rice is a staple crop in India, significantly contributing to food security and agricultural economies. However, rice cultivation faces substantial challenges due to various diseases that affect crop yield, particularly during the Kharif season (June-October), when warm and humid conditions favor pathogen proliferation. Manual disease identification by farmers is often inaccurate due to limited expertise, leading to improper disease management and yield losses. Recent advancements in deep learning, particularly Convolutional Neural Networks (CNNs), offer promising solutions for automated disease detection. This study proposes a CNN-based model leveraging Transfer Learning (TL) with a VGG16 architecture to classify rice leaf diseases efficiently. Due to the scarcity of publicly available datasets, we curated a custom dataset comprising images collected from rice fields and online sources. Despite the dataset’s limited size, transfer learning enabled robust feature extraction and improved classification performance. The model was trained and evaluated on real-world images, demonstrating high accuracy in identifying common rice leaf diseases. Our findings highlight the potential of deep learning in precision agriculture, providing farmers with an accessible tool for early disease detection and mitigation. The proposed approach can be extended to other crops, contributing to sustainable agricultural practices.

Introduction

Introduction

• Rice’s importance: A staple crop feeding over half the global population, especially in Asia and India.

• Challenge: Rice diseases (e.g., bacterial leaf blight, blast, sheath rot) cause major yield losses.

• Limitations of traditional diagnosis: Visual inspection is error-prone due to subjective judgment and similar-looking symptoms.

• Solution: Use AI and deep learning, particularly Convolutional Neural Networks (CNNs), for automated and accurate rice disease detection.

• Proposed system: A mobile-compatible, cloud-based platform where farmers can upload leaf images for real-time diagnosis and receive actionable treatment advice.

2. Literature Review

• Traditional ML approaches (SVMs, ANNs):

  • Required manual feature extraction.

  • Accuracy ranged from 78% to 83%.

• Deep learning with CNNs:

  • Automatically extracts features.

  • Achieved 92–96% accuracy across crops.

• Key advancements:

  • Lu et al.: Custom CNN for rice disease, 94.3% accuracy.

  • Atole & Park: Used transfer learning with AlexNet; 91.7% accuracy on limited data.

• Limitations of prior work:

  • Mostly tested in lab environments with ideal images.

  • Lacked real-world robustness and mobile deployment.

• This study’s contribution: A robust, real-world applicable system that maintains high accuracy, runs on mobile devices, and gives treatment advice tailored to local conditions.

3. Methodology

A. CNN Architecture

• Step-by-step breakdown:

  • Convolution → ReLU → Pooling → Flattening → Fully connected layers

• Supports automated feature learning from image data.

B. MobileNet

• Lightweight CNN optimized for mobile devices using depthwise separable convolutions.

• Balances accuracy with low computational cost.

C. Data Collection & Preprocessing

• Dataset: Custom + public (e.g., Kaggle); 6 classes including:

  • Healthy, Leaf Blast, Leaf Blight, Brown Spot, Leaf Scald, Narrow Brown Spot.

• Preprocessing steps:

  • Resize to 224×224

  • Normalize pixel values

  • Data augmentation (rotation, flips, zoom, brightness)

  • 80/20 train-validation split

D. Transfer Learning with VGG16

• Used VGG16 pre-trained on ImageNet:

  • Frozen convolution layers to retain feature extraction.

  • Custom classification head with:

    • Global Average Pooling

    • Dense (256 units, ReLU), Dropout (0.5)

    • Final Softmax (4 classes: Healthy, Leaf Blast, Leaf Blight, Brown Spot)

4. Model Training and Evaluation

• Setup:

  • 2,627 labeled images

  • Training with Adam optimizer, batch size = 32, learning rate = 0.0001

• Metrics used:

  • Accuracy, Precision, Recall, F1-Score

  • Confusion matrix and ROC curves

• Results:

  • High overall accuracy (≥90%)

  • Notable confusion:

    • 11% between Leaf Scald vs Narrow Brown Spot

    • 8% between Bacterial Blight vs Leaf Blast

5. Deployment & Practical Use

• Final product: A real-time, mobile-friendly application for farmers.

• Features:

  • Upload images → receive disease classification + treatment suggestions.

  • Designed to function in low-resource environments.

• Goal: Enable AI-powered precision farming to reduce crop loss and optimize pesticide use.

Conclusion

This research has effectively created and tested a deep learning model for precise rice leaf disease classification with transfer learning. In spite of the difficulty of using a small dataset of 1,509 training images, our VGG16 model fine-tuned to 92.46% classification accuracy on an independent test set of 647 images shows the power of transfer learning in surmounting data scarcity limitations that normally limit deep learning applications in agricultural settings. The strategic application of transfer learning using the VGG16 architecture was instrumental to our model\'s success. Trained from scratch on our limited dataset, the model did not deliver satisfactory results, emphasizing the value of utilizing pre-trained weights in order to successfully extract features. By careful fine-tuning, we were able to leverage the strong image recognition ability of VGG16 to the particular domain of rice disease detection without falling into the trap of overfitting. Our training approach included an early stopping strategy that stopped training after 25 epochs when both accuracy and loss metrics leveled off over training and validation sets. This not only optimized computational performance but also kept the model\'s generalization abilities robust. The convergence pattern we see indicates that our architecture reached a good balance between model complexity and training data available. These results have significant consequences for plant disease control in resource-poor environments. The fact that our method works as well as it does despite using relatively modest, human-annotated datasets suggests that automated disease diagnosis at high accuracy can be attained using relatively modest, well-constructed datasets. Possible directions for future work could involve extending the capabilities of the model to other classes of disease, optimizing the model for mobile operation to advantage field workers, and exploring multimodal methods combining vision data with other sensor modalities for improved diagnostic accuracy.

References

[1] T. Gupta, ”Plant leaf disease analysis using image processing technique with modified SVM-CS classifier,” Int. J. Eng. Manag. Technol, no. 5, pp. 11-17, 2017. [2] Y. Es-saady,T. El Massi,M. El Yassa,D. Mammass, and A. Benazoun, ”Automatic recognition of plant leaves diseases based on serial combination of two SVM classifiers,” International Conference on Electrical and Information Technologies (ICEIT) pp. 561-566, 2016. [3] P. B. Padol and A. A. Yadav, “SVM classifier based grape leaf disease detection,” Conference on Advances in Signal Processing (CASP), pp. 175-179, 2016. [4] L. Liu and G. Zhou, “Extraction of the rice leaf disease image based on BP neural network,” International Conference on Computational Intelligence and Software Engineering ,pp. 1-3,2009. [5] S. Arivazhagan and S.V. Ligi, “Mango Leaf Diseases Identification Using Convolutional Neural Network,” International Journal of Pure and Applied Mathematics, vol. 120,no. 6, pp. 11067-11079,2008. [6] B. Liu,Y. Zhang, D. He and Y. Li, “Identification of Apple Leaf Diseases Based on Deep Convolutional Neural Networks Symmetry,” X. X. &Suen, C. Y. A novel hybrid CNN–SVM classifier for recognizing handwritten digits. Pattern Recognition,vol. 45,pp. 1318–1325,2012. [7] Y. Lu, S. Yi,N. Zeng,Y. Liu, and Y. Zhang, “Identification of Rice Diseases Using Deep Convolutional Neural Networks,” Neurocomputing, 267, pp. 378-384,2017. [8] R. R. Atole, D. Park, “A Multiclass Deep Convolutional Neural Network Classifier for Detection of Common Rice Plant Anomalies,” International Journal Of Advanced Computer Science And Applications,vol. 9,no. 1,pp. 67–70,2018. [9] V. Singh, A. Misra, “Detection of Plant Leaf Diseases Using Image Segmentation and Soft Computing Techniques,” Information Processing in Agriculture,vol. 4 ,no. 1,pp. 41–49,2017. [10] P. KonstantinosFerentinos, “Deep Learning Models for Plant Disease Detection and Diagnosis,” Computers and Electronics in Agriculture ,vol. 145,pp. 311–318,2018. [11] S. P. Mohanty, D. P. Hughes, M. Salath´e, “Using Deep Learning for Image-Based Plant Disease Detection,” Frontiers in plant science ,vol. 7. [12] S. Xie, T. Yang, X. Wang, and Y. Lin, “Hyper-class augmented and regularized deep learning for fine-grained image classification,” IEEE Conference on Computer Vision and Pattern Recognition,CVPR2015,pp. 2645–2654,June 2015. [13] Y. Lu, S. Yi, N. Zeng, Y. Liu, Y. Zhang, ” Identification of rice diseases using deep convolutional neural networks”, Neurocomputing 267,pp. 378–384,2017. [14] K. Simonyan, A. Zisserman, “Very Deep Convolutional Networks for LargeScale Image Recognition”, Arxiv preprint arXiv: arXiv,pp. 1409- 1556,2015.

Copyright

Copyright © 2025 Mr. Balapuram Jayanth, Mr. A C Girish, Mr. D Kannan, Mr. D Charan Kumar, Mr. Pandreti Praveen, Dr. R Karunia Krishna Priya, Mr. N Vijaya Kumar, Mr. V Shaik Mohammad Shahil. 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.