Real-Time Soybean Crop Insect Classification Using Customized Deep Learning Models

One of the most challenging aspects of crop protection will always be flying, crawling, and chomping insects. With time, insects become susceptible to some insecticides, and hence, the right solutions (newly available insecticides) should reach the farmers at the right time. In this view, the study is aimed to provide a lightweight Artificial Intelligence (AI) driven system (can load in smartphone), which helps the farmer to identify, classify in real-time to control the target insects on chosen crops (soybean). The dataset (3809 images) was developed under MoU between IIIT-Naya Raipur (C.G.) and Indira Gandhi Krishi Vishwavidyalaya, Raipur, which consists of Eocanthecona Bug, Tobacco Caterpillar, Red Hairy Caterpillar, and Larva Spodoptera. The VGG16 and GoogLeNet deep learning models have been developed with the dataset and found decent performance as with VGG16 (validation accuracy: 97.78% and validation loss: 0.0852) and GoogLeNet (validation accuracy: 99.03% and validation loss: 69.5098). Finally, an Android-based app has been developed to facilitate real-time working.