•  
  •  
 

Abstract

Detecting and classifying insect pests is a critical challenge in agricultural pest management, as infestations can reduce crop yield and quality. This study introduces JPNet, a convolutional neural network (CNN) architecture that uses multi-layer feature fusion to detect and classify insect pests affecting jute crops. The architecture integrates complementary feature representations extracted at different network depths, preserving fine-grained visual characteristics alongside high-level semantic information. JPNet is evaluated on the JutePest dataset, which comprises approximately 6,460 RGB images spanning 17 pest classes. Preprocessing and data augmentation—including resizing, normalization, rotation, shifting, zooming, and flipping—improve the consistency and diversity of the training data. Performance is assessed using precision, recall, F1-score, accuracy, model size, parameter count, and training time, and compared with several established deep learning models. Under the evaluated experimental conditions, JPNet achieves an accuracy of 99.63\%, with precision, recall, and F1-score values of 0.99 each. The model contains approximately 50 million parameters and requires approximately 1.2 GPU hours for training under the reported experimental setup. These results suggest that multi-layer feature fusion enables effective classification of the 17 evaluated jute pest classes while maintaining competitive computational efficiency. However, evaluation on a curated dataset does not establish performance in field conditions or real-time applications. Further assessment using field-acquired images, independent datasets, unseen-domain testing, and systematic ablation studies is therefore needed to establish JPNet’s robustness and generalizability across diverse agricultural conditions. The framework provides a foundation for developing automated jute pest detection systems for future field applications.

Download

Share

COinS