Abstract:
About 40% of Indians are directly engaged in agriculture and 20% are indirectly engaged in agricultural jobs. The most widely grown crop worldwide is corn which is used in numerous agricultural products, including those that can be used to make biofuels, as well as in the food chain. In India, a large number of small-scale farmers rely on farming for both a living and for meeting their fundamental necessities. Conversely, corn crops are susceptible to illnesses that hamper crop yield and farmer income. Temperature fluctuations and unfavorable weather patterns cause the disease to spread. With the development of digital technology, the use of technology in farming and agricultural operations has widened. Farmers can use voluminous volumes of data regarding crop and soil conditions, climate change, and other environmental factors to guide their decisions about how to handle plants and animals through the use of machine-learning methods in agriculture. A modified deep transfer learning is implemented in this work, which classifies three major corn diseases and identifies healthy images among them. In this work, the most prevalent diseases were taken into account, including blight, gray leaf spot, and common rust. For forecasting the classes, Resnet-18, a type of convolutional neural network was deployed. The corn leaf image is provided as input and the transfer learning technique was established on Resnet-18 and the data was split extensively for multiple scenarios. It is classified into four classes and obtained a mean accuracy of 96% than the existing schemes.
