Generalising Across Different Crop Diseases with Deep Learning Model

Abstract

Global agriculture faces severe economic threats from plant diseases, necessitating automated diagnostic systems. However, standard deep learning models feature over-parameterized architectures that require prohibitive computational resources, limiting field deployment. This paper presents an efficient, lightweight sequential Convolutional Neural Network (CNN) optimized for rapid multi-class plant disease classification using the Kaggle New Plant Diseases Dataset. The proposed architecture streamlines feature extraction by applying a single max-pooling layer after every two 32-filter convolutional layers. To prevent overfitting, a dual dropout strategy (0.25 and 0.5) is integrated alongside a dense layer of 1,500 units. Trained over 10 epochs using the Adam optimizer (learning rate = 0.001) and varying batch sizes (32 to 512), the model demonstrated swift convergence, achieving a peak training accuracy of 98.15% and a validation accuracy of 95.87%. Evaluated across a test support of 70,295 images spanning 38 distinct crop classes, the network delivered an overall macro and weighted average of 1.00 for precision, recall, and F1-score. These results prove that an optimized, computationally economic sequential framework can match complex architectures, providing a viable solution for real-time edge deployment in precision agriculture.

Introduction

Plant diseases significantly threaten global agriculture and food security. Traditional visual inspection methods are slow, subjective, and often lead to delayed treatment. To overcome this, the study uses Convolutional Neural Networks (CNNs), which automatically extract features from plant leaf images and classify diseases with high accuracy. However, since very deep CNN models are computationally expensive and unsuitable for edge devices, the paper proposes a lightweight sequential CNN architecture designed for efficient real-time deployment in agricultural environments.

The proposed model is trained on the Kaggle Plant Disease dataset and consists of multiple convolutional layers with increasing feature maps, max-pooling for dimensionality reduction, and dropout layers to reduce overfitting. It includes a dense classification layer that outputs predictions across 38 plant disease classes. The model is trained using the Adam optimizer over 10 epochs.

Results show strong performance, with training accuracy reaching about 98% and validation accuracy around 96%, indicating good generalization with minimal overfitting. The classification report further confirms high precision, recall, and F1-scores across most classes, with near-perfect accuracy for many disease categories. Minor misclassifications occur in visually similar disease types.

Conclusion

This paper successfully developed a lightweight, sequential Convolutional Neural Network (CNN) tailored for rapid and precise plant leaf disease detection using the Kaggle New Plant Diseases Dataset. By strategically utilizing a localized architecture featuring 32-filter convolutional layers paired with a max-pooling layer after every two convolutions, the model significantly optimized computational efficiency. The implementation of a dual dropout mechanism (0.25 and 0.5) proved highly effective in mitigating overfitting, enabling the model to bridge the gap between high accuracy and minimal parameter dependency. Experimental results underscore the success of this approach, with the model achieving a peak validation accuracy of 95.87% within just 10 epochs. Furthermore, the model achieved a macro and weighted average of 1.00 across precision, recall, and F1-score metrics evaluated over a test bank of 70,295 images. These findings prove that heavy, deeply layered neural frameworks are not strictly necessary to achieve high-fidelity phytopathological classification. Ultimately, this study delivers an accessible, robust, and computationally economic framework. In future work, this sequential model can be integrated into mobile applications or low-power edge computing devices, providing farmers with a viable, real-time diagnostic tool to support precision agriculture and reduce crop yield loss.

References

[1] S. K. Mishra, K. V. D. Jakkinapalli, N. Chanda, R. Ganta, N. Linga, and P. Charupalli, \"Plant Disease Detection and Fertilizer Recommendation using CNN,\" in 2024 First International Conference on Innovations in Communications, Electrical and Computer Engineering (ICICEC), 2024, pp. 1-7, doi: 10.1109/icicec62498.2024.10808203. [2] M. H. Saleem, J. Potgieter, and K. M. Arif, \"Plant Disease Detection and Classification by Deep Learning,\" Plants, vol. 8, no. 11, p. 468, 2019, doi: 10.3390/plants8110468. [3] Y. Toda and F. Okura, \"How Convolutional Neural Networks Diagnose Plant Disease,\" Plant Phenomics, vol. 2019, p. 9237136, 2019, doi: 10.34133/2019/9237136. [4] Nyawose, T. (n.d.). A review on the detection of plant disease using machine learning and deep learning approaches. PMC. https://pmc.ncbi.nlm.nih.gov/articles/PMC12565507/ Cited by: 36 [5] Kumar, R., Chug, A., & Singh, A. P. (2023). An efficient plant leaf disease detection model using Shallow-ConvNet. Applied Ecology and Environmental Research, 21(4), 3193–3211. https://doi.org/10.15666/aeer/2104_31933211 Cited by: 5 [6] Tugrul, B., Elfatimi, E., & Eryigit, R. (2022). Convolutional neural networks in detection of plant leaf diseases: A review. Agriculture, 12(8), 1192. https://doi.org/10.3390/agriculture12081192 Cited by: 332 [7] Nyawose, T. (n.d.). A review on the detection of plant disease using machine learning and deep learning approaches. PMC. https://pmc.ncbi.nlm.nih.gov/articles/PMC12565507/ Cited by: 36 [8] Dhaka, V. S., Meena, S. V., Rani, G., Sinwar, D., Kavita, K., Ijaz, M. F., & Wo?niak, M. (2021). A survey of deep convolutional neural networks applied for prediction of plant leaf diseases. Sensors, 21(14), 4749. https://doi.org/10.3390/s21144749 Cited by: 495 [9] Pandian, A. (n.d.). An improved deep residual convolutional neural network for plant leaf disease detection. PMC. https://pmc.ncbi.nlm.nih.gov/articles/PMC9492343/ Cited by: 48

Copyright

Copyright © 2026 Shubham Verma, Anita Pal. 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.