JuteNet: An Intelligent Approach for Jute Pest Recognition Using Residual Network with Hybrid Attention Module

The harm of pests and diseases to crops is often irreversible. In traditional agriculture, farmers' identification of pests and diseases primarily relies on experience, which has defects such as low efficiency and high misjudgment rate. Hiring experienced experts will significantly increase the cost of investment. Given the above problems, this study proposes an improved residual network for jute pest identification. Based on the ResNet18 model, a hybrid attention module is embedded to construct a new convolutional neural network, JuteNet. The experimental results show that the recognition accuracy of the JuteNet network is 93.24%, which is 3.47% higher than that of the ResNet18 network. By introducing an appropriate amount of attention module, the recognition accuracy of the network can be improved without significantly increasing the training cost of the model. In addition, the network's superiority is further demonstrated by ablation experiments and comparison with other models. This experiment verifies the feasibility of a deep-learning algorithm for plant pest identification. It provides a more efficient and convenient solution for intelligent and accurate classification of plant pest detection.