Modified RESNET50 with Attention Module for Detection and Classification of Pests in Vegetable Crops

Abstract

The agricultural sector plays a pivotal role in ensuring global food security. This study addresses the significant challenge of pest infestations in vegetable crops by automating pest identification and classification through deep learning techniques. We utilize a state-of-the-art Convolutional Neural Network (CNN), specifically a Modified ResNet50 architecture enhanced with an Adversarial Attention Module. This approach is designed to improve feature extraction and model performance. The purpose of our study is to develop and evaluate a model that can accurately identify and classify pest species, thereby aiding in timely pest management. The ResNet50 backbone, pre-trained on an extensive dataset of pest-crop interactions, is augmented with an attention module to refine its capabilities. Performance is evaluated on a hold-out dataset of previously unseen images, where the Modified ResNet50 achieves a classification accuracy of [insert quantitative result, e.g., 92%]. Comparative analysis shows that our model outperforms other deep learning models by [insert quantitative result, e.g., 5%] in precision and recall metrics. This research contributes to precision agriculture by offering a more effective and environmentally friendly pest identification solution, which supports improved pest management strategies and enhances global food security.