IoT based Soyabean Pest Classification using Transfer Learning and Explainable AI

Soybean crops face significant susceptibility to pests like red hair caterpillars, leading to diminished crop yield and an overall reduction in crop quality. With this perspective, our work aims to provide a lightweight model powered by Artificial Intelligence (AI) that assists farmers in real-time identification and classification of target insects on selected crops (soybean). In this paper a model has been proposed with the Inception-ResNet-V2 framework, which showed respectable performance (99.54% validation accuracy and 0.0103 validation loss). Under an agreement between IIIT-Naya Raipur (C.G.) and Indira Gandhi Krishi Vishwavidyalaya, Raipur, the dataset (3809 images) comprising the Eocanthecona Bug, Tobacco Caterpillar, Red Hairy Caterpillar, and Larva Spodoptera was used for comparative study in the proposed model. The proposed work is found appropriate for research purposes because it uses Explainable AI (Artificial Intelligence) to validate the classification process. The proposed model is also integrated and implemented with the on-board Raspberry Pi, making it more interactive and useful in the agricultural domain.