Rice Diseases Pests and Nutritional Deficiency Classification using Convolutional Neural Network

The major cereals crops cultivated in India is wheat, pearl millet, rice, sorghum and maize they accounts for around 52% of the total crop cultivation. India being the second largest producer of rice after China exports rice to over 150 countries in the world. In order to increase the productivity to the fullest the crop field must be free from the diseases (fungal and bacterial), viruses, pests and nutritional deficiencies. As the rice crop gets affected due to the diseases like brown spot, sheath blight, blast, sheath rot and etc. if not detected can kill the rice leaf. There are some major pests of the rice crop attacks every portion of plant stem bores, leaf hoppers, grain sucking insects, root feeders, defoliators and etc. These pests usually disturb the plant growth cycle by sucking the plant sap, reducing root system, damaging shoots and reduces photosynthetic capacity thereby reducing yields. Also some symptoms on leaf appears when the plant suffers from nutritional deficiencies: micro nutrients required in greater quantity whereas macro-nutrients required in small quantities. In order to increase the crop exports and to feed the growing population the crop field must be free from all the pests, diseases and deficiencies. In this paper we proposed a model which detects the rice crop pests, diseases and primary nutritional deficiencies with good accuracy in the real time. In our experiments we have trained multiple transfer learning algorithms InceptionV3, VGG19, Xception, InceptionResNetV2 and also a CNN model without transfer learning. The models are trained on the datasets from the kaggle database and the accuracy of different models are compared and finally with InceptionV3 and InceptionResNetV2 model we achieved a best validation accuracy of 95.56%.