Abstract:
In the modern agricultural world, early detection of plant diseases is critical for
growing healthy crop and reducing losses. Old and traditional methods to identify a
plant disease are often time consuming and requires a expert knowledge, which makes
them less accessible to regular farmers and people. This project focuses on the
development of a mobile based disease detection system that uses an advanced deep
vision techniques to accurately identify diseases in vegetable plants very specifically
Tomato, and Bell Pepper (Capsicum) crops. This system focuses on key leaf
Potato,
diseases, including Late Blight and Early Blight of Potatoes, Tomato Mosaic Virus of
We studied various Convolutional
Tomatoes, and Bacterial Spot of Bell Peppers.
Neural Network (CNN) architectures, like MobileNetV2, VGG16, ResNet50,
EfficientNetBO, and InceptionV2 to select model that is suitable for our project and
gain good knowledge about the overall model accuracy, efficiency, and behaviour.
This deep analysis significantly contributed to our understanding of machine learning
and deep vision techniques. Based on our findings, we used MobileNetV2 model
which is suitable for mobile application. Our model achieved an accuracy of 95.5%.
different Environmental conditions.
cameras.
This allows plant disease detection from many
This model communicates with the custom mobile application we developed through
fastAPI, which allows users to scan plant leaves for diseases using their smartphone
The mobile app, along with the FastAPI based backend, is deployed
cloud service to make sure smooth, reliable, and efficient system performance
