Comparison of Deep Transfer Learning Models for Pest Image Classification in Agriculture

Abstract

In the field of agriculture, crops are susceptible to attacks by pests. Accurately identifying and classifying crop pests, especially in their early stages of growth, is a significant challenge. Convolutional neural networks have become effective instruments for agricultural pest classification because of their capacity to extract and learn intricate information from photos. This study explores the use of transfer learning methods to compare efficient models with complex models to improve the effectiveness of pest and disease classification. The purpose of this study is to compare efficient models with complicated models using transfer learning techniques in order to increase the accuracy of pest and disease classification. By conducting experiments on the same pest and disease dataset, this research compares their performance in terms of accuracy, model size, computational resource consumption, and practical feasibility in real-world applications. The research results clearly demonstrate the performance comparison between different models, with efficient models excelling in some aspects while large models have advantages in others. This study contributes to better understanding for decision-makers in the agricultural sector on how to choose the appropriate deep learning models to enhance crop protection and increase yields.