Diagnosis Of Cassava Leaf Diseases and Classification Using Deep Learning Techniques

Abstract

The plants disease diagnosis is very challenging research in the field of agriculture. Cassava is a second most provider of carbohydrates in Africa. It is a key food for people of Africa in very harsh conditions. According to United Nations (FAO) almost eighty percent farmers of sub Saharan Africa are growing cassava roots, but due to a variety of viral diseases the production of cassava is very low from last two years. With the help of data science, it is possible to diagnose and classify these types of viral diseases. Existing methods of disease detection require farmers to solicit the help of government-funded agricultural experts to visually inspect and diagnose the plants. Moreover, this process is labor-intensive, time taken, costly and impacting the production and supply cycle. As an added challenge, effective solutions for farmers must perform well under significant constraints since African farmers may only have access to mobile-quality cameras with low-bandwidth. The dataset which we use in this research is taken from Kaggle competition 2020. Dataset contains 21397 images of cassava plants which belongs to five different classes i.e., Cassava Bacterial Blight, Cassava Brown Streak Disease, Cassava Green Mottle, Cassava Mosaic Disease and Healthy leaves. In this work we have used augmentation technique to increase the samples for classification and balancing the uneven distribution of data for all classes and used deep learning model efficiennetB3 for identification classification of diseases and got 83.03% overall accuracy on test dataset with more than 90% individual accuracy of each class. We have developed a graphical user interface for using the model in more efficient way with the aim to help the industry for prediction of diseases during its initial stages.