Automatic Detection of Cassava Leaf Disease using Transfer Learning Model

Africa is the world's leading producer of cassava crops. It is also extensively distributed in Asia and Latin America. Cassava is produced and exported in large quantities from Thailand, which is the world's top producer of it. It may grow and function well in poor soil conditions. Cassava may be grown successfully in any climate. In addition to the nutrients present in rice, it has various health benefits. It was a significant source of calories and carbs. Cassava production has been reduced since 2016 owing to cassava disease. Cassava disease is the principal cause of decreased cassava yield and a direct influence on farmer income. A lot of techniques were employed in the recent years but there is scope for improvement in performance parameters. In this research, a pre-training model based on deep learning is used to detect cassava leaf disease. The illness is classified into five categories Cassava Bacterial Blight (CBB), Cassava Brown Steak Disease (CBSD), Cassava Green Mite (CGM), Cassava Mosaic Disease (CMD) and Healthy. The deep learning model used in this study is densenet169. The model is implemented using a collection of 21397 images having 5 classes of cassava leaf. The model has been evaluated in terms of performance parameters such as loss, accuracy, specificity, sensitivity.