Classification for Crop Pest on U-SegNet

Abstract

As a long-standing issue, pests and illnesses in agriculture consistently cause large annual yield decreases. It's no secret that deep learning (DL) models are quite effective in facial recognition, therefore many people in the agricultural industry are curious about them. Since the algorithms used in existing methods of pest identification are sophisticated and there is a shortage of relevant data, the methods' accuracy in recognising and categorizing pests is low. When insects are misidentified as another type of pest, the wrong pesticides may be used, negatively affecting both crop productivity and the environment. Pre-trained deep learning architectures like Unet and ResNet were used to compare the proposed model to the suggested U-SegNet for insect categorization. This study also enhances the data to stop the network from becoming overly specialized. The accuracy of the proposed model has been analyzed by examining the impact of hyperparameters. The highest possible accurate classification rate has been accomplished at 93.54%.