Zero Hunger - Crop Disease Detection using Computer Vision

Abstract

The essential economic contribution of agriculture to developing nations helps maintain food security which serves as the foundation of their economic systems. Farmers face a major difficulty because they need to identify crop diseases at an early stage through accurate methods because these diseases will cause major crop losses if they remain undetected. The existing methods for detecting diseases require experts to conduct manual inspections which develop into a process that consumes excessive time and incurs high costs while becoming unsuitable for implementation in extensive agricultural operations. The project proposes a Crop Disease Detection System which uses Deep Learning techniques as a solution to these existing challenges while supporting the Sustainable Development Goals 2 Zero Hunger. The system uses Convolutional Neural Networks (CNNs) for automatic detection and classification of crop diseases through its analysis of leaf images. The dataset includes images of healthy and diseased crop leaves which researchers obtained from both public databases and real-world environments. The images undergo preprocessing through three steps which include resizing and normalization and augmentation to achieve model accuracy and robustness improvements. The proposed solution supports sustainable farming through its early disease detection capabilities and precision agriculture functions which lead to better crop yields and decreased food shortages. The project shows how deep learning functions as an effective agricultural tool while demonstrating how artificial intelligence enables sustainable solutions which help achieve the zero-hunger objective.

Code and data availability

The paper uses the public PlantVillage dataset and a custom CNN, but provides no authors' code, trained model, or paper-specific data deposit; no availability statements or public repository URLs for the authors' assets appear in any block. PlantVillage is a cited prior public dataset, not a paper-specific asset.

No evidence-backed public reproduction asset is currently recorded.