| dc.description.abstract | Timely detection of plant diseases is vital for social and global benefits because it
ensures food security by preventing significant crop losses, thus maintaining stable food
supplies for a growing population. Early disease management supports economic
stability for farmers by avoiding devastating financial setbacks. It also promotes
environmental sustainability by reducing the need for extensive chemical treatments,
conserving natural resources, and protecting ecosystems. Overall, timely plant disease
detection is crucial for fostering a resilient, sustainable, and healthy global community.
The research question is to find ways to improve the effectiveness and accuracy of plant
leaf disease detection using recent deep learning approaches emphasizing
effective segmentation, feature extraction, and classification techniques. The primary
research objective is to assess modern deep learning models to be effectively utilized
for the timely and accurate detection of plant diseases to enhance crop production.
To answer the research question, convolutional neural networks (CNNs) are employed
for classifying diseases across multiple crops, including cotton, apples, and maize for
more generalized solution. This study aims to provide an in-depth investigation of
cutting-edge plant leaf disease detection models, including their approach, strengths,
constraints, and essential computational strategies such as segmentation, feature
extraction, and classification. By training and evaluating multiple pre-trained CNN
models, the study assesses their ability in identifying and classifying plant diseases from
leaf images, ultimately providing valuable insights into how these models can be used
to improve plant disease detection approaches. The study trained several pre-trained
CNN architectures with fine-tuning on cotton disease dataset and shows the
effectiveness of MobileNet-V2 and Inception-V3 models in terms of accuracy,
precision, F1-score and recall. | en_US |
