Copyright © 2026 Authors retain the copyright of this article. This article 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.

@article{199595,
        author = {Arunendra Singh},
        title = {A Unified Deep Learning Framework for Plant Disease and Insect Pest Classification},
        journal = {International Journal of Innovative Research in Technology},
        year = {2026},
        volume = {12},
        number = {12},
        pages = {2479-2486},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=199595},
        abstract = {Plant diseases and insect pests are major threats to global food security, causing up to 40% of annual crop losses amounting to over $220 billion. Early and accurate identification of these biotic stresses is essential for effective crop protection and integrated pest management (IPM). Traditional manual inspection by agricultural experts is time-consuming, subjective, and impractical for large-scale farms. Recent advances in deep learning, particularly convolutional neural networks (CNNs), have shown remarkable potential for automated image-based diagnosis. However, most existing deep learning models focus exclusively on either plant diseases or insect pests, requiring separate systems for comprehensive crop health monitoring.
This paper proposes a unified lightweight deep learning framework based on MobileNetV2 for simultaneous classification of 58 distinct categories, including 38 common plant diseases and 20 harmful insect pests. The dataset comprises over 53,000 leaf images, split into training (80%) and validation (20%). Transfer learning is employed with a frozen MobileNetV2 base pre-trained on ImageNet. Custom layers include global average pooling, a dense layer with 256 units and ReLU activation, dropout (0.5) for regularization, and a softmax output layer for 58 classes. Online data augmentation (random rotation, zoom, horizontal flip, brightness adjustment) improves generalization and reduces overfitting.
The proposed model achieves a validation accuracy of 88.19% and validation loss of 0.5007 after 10 epochs. Per-category analysis shows 89.2% average accuracy for plant diseases and 86.4% for insect pests. The compact architecture contains only 2.6 million parameters, making it suitable for deployment on resource-constrained edge devices and mobile platforms. The model can be quantized to under 2 MB for real-time inference (under 200 ms per image). This unified approach eliminates the need for separate disease and pest detection systems, offering a practical, cost-effective tool for farmers and agronomists. Future work includes field testing, model pruning, and mobile app deployment.},
        keywords = {Plant disease detection, insect pest classification, MobileNetV2, transfer learning, deep learning, precision agriculture, integrated pest management, lightweight CNN, edge computing.},
        month = {May},
        }