Paddy Pest Identification and Pesticide Recommendation Using Fine-Tuned ResNet-34 and Rule-Based Mapping

Paddy or rice is the second highly consumed staple food in the world. In Asian countries especially in China and India, Paddy is the primary source of food. The most common problem faced by paddy cultivators is pest infestation. In this case, early detection of pest type and proper pesticide recommendation System becomes very necessary. In this study, a heavy learning framework for the classification of paddy pests is developed through a fine-tuned ResNet-34 model along with a rule-based pesticide recommendation approach. Data containing images of six different types of paddy pests were used to train and test upon the proposed model. Validation accuracies of 89.45% were achieved by the system, portraying great reliability and automation potential which could help in higher accuracy in pest disease identification. This system can reduce manual work, increase reliability and automation.