Onion Disease Prediction Using Deep Learning

Abstract

Now a days in Bangladesh, onion disease is one of the most serious agricultural problems. As a result, most people in our country value onion care. While recent advances in computer vision have made object detection from images much easier, automatically classifying onions with computer vision remains a difficult task due to similarities between different types and factors such as their location (e.g., stacked) or lighting conditions. A framework for classifying onion diseases can be useful in a variety of fields, including autonomous agricultural robotics and the development of mobile applications for detecting specific onion diseases on the market. We tested two different models for fruit detection that used deep convolutional neural network (DCCN) techniques in this paper, and based on our training results, we proposed an efficient model. Dense-net-201 and AlexNet were used to train with endemic Bangladeshi fruits. Images of onion diseases from six different disease classes were included in our dataset. The dataset was split into two parts: 80% for training and 20% for research. For easier preparation, the training dataset was augmented and expanded. With our own dataset, we achieved a high accuracy rate of 93.83 % with the AlexNet model.