Smart Detection of Pumpkin Leaf Diseases Using MobileNetV3: A Step Toward Healthier Crops
Richa Brar, Richa Chandel, Amandeep Kaur, Monika Sharma, Mani Goyal, Pratham Kaushik
Abstract
The research employs MobileNetV3-Large, a light and efficient deep learning model, to classify pumpkin leaf diseases from a public domain dataset. Various data preprocessing techniques, such as bilateral filtering, histogram equalization, and adaptive thresholding, improved image quality and classification accuracy. The data were partitioned into 80 % train, 10 % validation, and 10 % testing for efficient learning. The model achieved 88 % accuracy with good precision, recall, and F1-score, indicating excellent performance. The confusion matrix revealed small misclassifications in Bacterial Leaf Spot with areas of improvement. This research provides a cost-efficient and scalable deep learning solution to plant disease detection automation. Future research might explore enhanced feature extraction, hyperspectral imaging, and IoT-based real-time deployment to improve accuracy and usability in smart agricultural systems.