Bangladeshi Crop Disease Detection Using Convolutional Neural Network

Self-sufficiency in food production is a vital survival tool for a country like Bangladesh in a deregulated world. However, crop diseases pose a significant obstacle to achieving self-sufficiency in food production in this small, agriculture dependent economy. The primary crops in Bangladesh include corn, rice, potato, and wheat. Every year, approximately 10% to 15 % of crops are lost due to insect infestations. Additionally, the lack of accurate diagnostic tools results in the unnecessary and excessive use of pesticides, which not only increases the cost of production but also contaminates the food supply. But nowadays, accurate and fast diagnosis is possible with the help of deep learning like Convolutional Neural Network (CNN). The primary goal of this paper is to accurately identify diseases in the main crops of Bangladesh, including corn, rice, potato, and wheat. This is achieved by utilizing EfficientNet-B3 as the backbone network. The model achieved an accuracy of 99% for corn, 92% for rice, and 100% for both potato and wheat, surpassing the performance of previous models.