Agro Disease Finder System of Cauliflowers Using Convolutional Neural Network

Abstract

In an underdeveloped country in South Asia like Bangladesh, every year, many farmers have suffered significant losses as a result of the cauliflowers illness. Farmers have no advanced knowledge of infection detection and mitigation strategy. They can understand the infections after the cauliflowers had already been damaged and in that, they have nothing to do in this matter. Many farmers are now afraid to take efforts to grow cauliflowers because of the loss of production. In this regard, we conducted a study using the improvement of artificial intelligence technology to identify and categorize cauliflowers sickness. We provide an online computer vision technology for developing an Agro disease finder system (AdFS) that analyzes a cauliflower picture recorded with a mobile or portable device and identifies illnesses, allowing faraway farmers to treat the problem. Firstly, we made a dataset with the help of agricultural expertise. They help us to identify the unhealthy cauliflower by examining it in their laboratory after that we took the photo and made this dataset for our research work. Our dataset contains 444 infected cauliflower images which categorize into four types of diseases. According to the TensorFlow and Keras APIs, we utilized the CNN model. This model is dependable and completely linked with all segmentation accomplished. The whole procedure is dependent on deep learning and the technique we use here is called the transfer learning technique. For accomplishing this problem, we used three state-ofthe-art algorithms and that is VGG19, VGG16 and ResNet50.