CONFERENCE / ICCAIS-2026

The impact of plant leaf diseases on crops can result in significant losses in agricultural productivity when left unchecked or not detected early on. Methods currently utilized to detect diseases on plants include manual inspection and personal expert insight, both of which can be extremely labour-intensive for large farms and take extended periods to complete. This paper presents an automated plant leaf disease detection system using deep learning and transfer learning techniques. The proposed approach employs the EfficientNetB0 convolutional neural network model to classify plant leaf diseases from images. A publicly available plant disease dataset containing approximately 27,000 images across 20 disease classes and 8 plant species was used for training and evaluation. Image preprocessing and data augmentation techniques were applied to improve model generalization. The experimental results show that the base model achieved an accuracy of 98%, which was further improved to 99.23% through fine-tuning. To demonstrate real-world applicability, the trained model was integrated into a mobile application developed using the Flutter framework, enabling users to capture or upload leaf images and obtain disease predictions along with recommended treatment and preventive actions. The proposed system supports early disease detection, reduces dependency on manual monitoring, and provides an effective decision-support tool for smart and sustainable agriculture.