Abstract:
This thesis extensively evaluates deep learning models for classifying plant diseases
and their effects on agriculture, society, and the environment. Basic CNN, Modified
AlexNet, EfficientNet B0, and EfficientNet B4 are compared for their ability to
diagnose and treat agricultural diseases. The models were very accurate, with the
EfficientNet B4 model scoring 99.99%. The Modified AlexNet and EfficientNet B0
models followed with 99.6% accuracy. CNN's 99.5% accuracy was impressive. The
models also had modest loss values, suggesting good learning and training. With its
accuracy, recall, and F1-score measures, EfficientNet B4 reliably classified all disease
categories better than any other model. The models' capacity to transform agriculture,
enhance crop management, and secure global food security has a major influence on
society. The research also examines how deep learning models reduce chemical
pesticide consumption and promote sustainable farming. Ethical considerations include
data privacy, algorithmic bias, transparency, and responsibility emphasize the need for
responsible and fair deployment. A sustainability strategy plans the long-term viability
and inclusiveness of agricultural deep learning models. Future research should address
dataset limits, examine a variety of data sources, improve model performance, and
evaluate ethical issues, according to the thesis. This study highlights the potential of
deep learning models in agriculture. It stresses ethical and sustainable methods while
using these models to solve social issues and encourage environmental stewardship.
