Plant Leaf Disease Detection by Using MobileNet and Inception

Detection of plant leaf disease in early stage is a very needy task in the field of agriculture because it helps to reduce the crop loss rate and increases the production of food. The purpose is to utilize Inception and MobileNet for disease detection as a conventional technique to detect plant disease in leaf manually which is cumbersome, subjective, faulty prone and expensive. Despite their efficiency in machine learning techniques, these are generally requiring a lot of reengineering on features and failing to deal with the large heterogeneous data set. In this project, we use MobileNet and Inception individually to develop a deep learning based approach to detect the plant leaf disease such as Apple scab, Black rot, Bacterial spot, Early blight, and Powdery mildew, etc. In the past times, MobileNet and Inception (advanced CNN models) did the trick of image classification tasks with their enormous ability to automatically detect a disease from the leaf image with least preprocessing only. It can be inferred that MobileNet was accurately capable of achieving an algorithm accuracy of 90% for real time in the field detection application for its computation efficiency and the best suitability for mobile devices. As inception architecture is significantly deeper than the network architecture used, more complex features are captured and the model accuracy achieved is 88%. It eliminates the manual feature extraction and directly learns from the raw images provided to it. The given advantage is to increase the production and detection of the plant leaf diseases early.