ABSTRACT: Plant diseases can reduce crop quality and yield when they are not identified at an early stage. This paper presents a practical and easy-to-understand system for plant leaf disease classification and segmentation. The proposed system accepts a photograph of a single plant leaf and uses a Convolutional Neural Network (CNN) to predict the probable disease class. A second image-processing stage based on K-Means clustering or Otsu’s thresholding is used to highlight the suspected diseased region and estimate the percentage of the leaf that appears affected. The system uses the publicly available PlantVillage dataset and focuses initially on crops such as tomato, potato, and grape. Images are resized, normalized, and augmented before CNN training. The classification model uses the Adam optimizer and categorical cross-entropy loss. The complete solution is designed with free and open-source tools including Python, TensorFlow/Keras, OpenCV, scikit-image, and Flask. The aim is not to replace an agricultural expert, but to provide a simple first-level screening and visual assistance tool that tells the user both what disease the model predicts and which part of the leaf appears to be affected.
KEYWORDS: Plant disease detection, CNN, image classification, image segmentation, K-Means, Otsu thresholding, PlantVillage, Flask, deep learning.
PLANT LEAF DISEASE CLASSIFICATION AND SEGMENTATION USING CNN AND IMAGE PROCESSING
MRS. MANALI CHURI, MR. SUJAL SHINDE