Study on the automatic detection of pepper TSWV disease based on ResNet

Abstract

This study focuses on the development of a binary classification model for Tomato Spotted Wilt Virus (TSWV) recognition in peppers. A diverse dataset is created for model training, which incorporates TSWV-infected pepper imagery with varying infection severity and regions of different sizes and resolution. The model is trained based on ResNet-18 architecture, resulting in excellent performance of metrics with scores reaching 0.990. Thus, the developed model may have applicational potential for pepper TSWV recognition in real-life conditions.