TOWARD SMART AGRICULTURE: CITRUS FRUIT DISEASE DETECTION USING DEEP LEARNING

Abstract

Agriculture plays a vital role in Pakistan’s economy, with citrus being one of the most widely cultivated fruits. However, citrus diseases such as scab, canker, and blight cause substantial losses by reducing both the quality and yield of the fruit. Traditional methods for disease identification rely on expert inspection, which is often time-consuming, costly, and impractical for large-scale monitoring. To overcome these limitations, this study presents a deep learning–based approach for automated detection of citrus canker. A dedicated dataset was collected using a Nikon D-5300 DSLR camera, pre-processed, and annotated to ensure accurate model training. We employed Faster R-CNN integrated with InceptionV3 and ResNet50 networks to detect canker-affected regions with high precision. The proposed method not only accelerates disease diagnosis but also reduces dependency on human expertise. By merging agriculture with advanced deep learning, this work contributes towards efficient, timely, and cost-effective disease management for citrus production.