Lychee Leaf Disease Detection Using Computer Vision and Deep CNN

Abstract

There has been a consistent increase in the demand for litchi in South Asia. Due to the fact that it is a fruit that is only available during specific times of the year, it is extremely attractive, particularly in Bangladesh. In addition to that, it possesses a flavor that is distinct from one another. At the field level, the cultivation of this crop is continuously increasing due to the fact that, in comparison to other crops, it has a comparatively cheap investment required for cultivation. However, the most important problem at hand is the fact that it is related with a number of ailments. For the most part, the leaves that are found on the litchi plant. Within the scope of this work, Deep CNN is utilized to effectively identify two widespread disorders that affect litchi leaves. Through my own efforts, I was able to physically gather over 7,000 diseased leaves from a number of litchi gardens and photograph them. The collection of such enormous amounts of data by manual means constituted a huge challenge. In the course of this inquiry, I learned three different models by the application of Deep Convolutional Neural Networks (CNN). A perfect performance was reached by the third model as VGG16 which is 0.99%. Both twining blight and life blight can be easily identified on fresh leaves thanks to its ability to detect them.