Deep Learning for Cotton Pest Detection: Comparative Analysis of CNN Architectures

Cotton is an important crop in Maharashtra, India. Heavy losses of cotton crop due to the pest attack. This study was carried out during the cropping season of 2022 and 2023 at Innovation Centre of Priyadarshini College of Engineering, Nagpur, and Maharashtra, India. The National Bureau of Agricultural Insect Resources (NBAIR) dataset, containing 40 classes of insect images, was utilized with integrating some real time cotton pest image dataset which were collected from farm nearby Umred village, Nagpur, Maharashtra. The study employed Convolutional Neural Network (CNN) architectures such as Alex Net, VGGNet-16, VGG-19, model and proposed new Convolutional Neural Network model for pest identification on cotton plants using Python software in google colab notebook with supporting libraries of Keras, Tensor flow and Open CV. Comparative experimental results showed accuracy rates of 96.75%, 97.23%, 95.97% and 97.87% respectively demonstrating the method's effectiveness in accurately identifying pests on cotton plants.