CROP LEAF DISEASE SEVERITY CLASSIFICATION USING PLANTVILLAGE IMAGE FEATURE EXTRACTION

Abstract

Agriculture plays a vital role in global food security, yet crop diseases remain a major challenge
affecting productivity and quality. Early and accurate identification of plant diseases is essential
to reduce crop losses and ensure sustainable agricultural practices. Traditional disease detection
methods rely heavily on manual inspection by experts, which is time-consuming, costly, and
often impractical for large-scale farming. With the rapid advancement of Artificial Intelligence
(AI) and computer vision technologies, automated crop disease detection using leaf images has
emerged as an effective and scalable solution.
This project proposes an Ai driven crop disease detection that utilizes digital images of plant
leaves to identify and classify diseases at an early stage. High-resolution leaf images are
collected from publicly available agricultural datasets and real-world sources. These images
undergo preprocessing steps such as resizing, noise reduction, and normalization to enhance
image quality and improve model performance. The processed images are then used to train deep
learning models capable of extracting relevant visual features related to disease patterns.
Convolutional Neural Networks (CNNs) and transfer learning techniques form the core of the
proposed system. Pretrained models such as VGG, Reset, or Mobile Net are fine-tuned on cropspecific
datasets to improve classification accuracy while reducing training time and
computational cost. The trained model learns to differentiate between healthy and diseased
leaves and further categorizes diseases based on visible symptoms such as spots, discoloration,
and texture variations. The system is designed to provide accurate disease prediction along with
confidence scores, enabling farmers to make informed decisions regarding crop treatment. By
integrating the model into a user-friendly web or mobile application, farmers can simply upload
leaf images and receive instant diagnostic results. This approach minimizes dependency on
agricultural experts and allows timely intervention, thereby reducing excessive pesticide usage