Plant Disease Detection

Plant Disease Detection Using Digital Image Processing and Deep Learning

Challenge

Plant diseases severely impact crop productivity, profitability, and food security—especially for fruit crops like apples, where infections spread rapidly and often go unnoticed in early stages. Traditional disease diagnosis depends on expert intervention, manual inspection, and visual judgment, making it slow, inconsistent, and inaccessible for farmers with limited agricultural support. A scalable, automated system was needed to diagnose diseases quickly and accurately using image-based inputs.

Our Solution

To address these challenges, I developed an image processing and deep learning–based disease detection system focused on apple plant leaves.

Key components of the solution included:

• Deep Learning Architecture:

  A hybrid model combining YOLO for detection and CNN classifiers for disease categorization, enabling precise localization of affected regions followed by high-accuracy classification.

• Digital Image Processing Pipeline:

  Preprocessing steps such as color normalization, noise filtering, leaf segmentation, and feature extraction were implemented to improve model robustness and reduce false detections.

• Custom Detection GUI:

  A simple, intuitive graphical interface was created to allow users to upload leaf images and instantly receive disease predictions, confidence scores, and highlighted affected areas.

• Dataset Curation & Training:

  A high-quality dataset of apple leaf diseases was prepared and used to train both models, optimizing performance across multiple disease types with strong accuracy metrics.

This standalone prototype demonstrated how AI and image processing can make plant health assessment faster and more accessible.

Impact

The system significantly streamlines the disease monitoring process, enabling early detection and reducing crop loss through timely intervention. Its GUI makes it practical for researchers, agricultural advisors, and even non-technical users. The approach also highlights the potential for deployment in mobile applications or integrated farm management systems.

Outcome

The work was published a in the Grenze International Journal of Engineering and Technology (GIJET) and was presented at the Hinweis RTET Conference. The project validated the effectiveness of combining detection and classification networks for agricultural diagnostics and laid groundwork for future field-ready AI tools.