基于自适应权重优化的多任务深度学习模型在甘蔗病害识别中的应用

Research on the application of multi-task deep learning model based on adaptive weight optimization in sugarcane disease identification

针对农业领域中甘蔗病害识别的准确率和任务间平衡的问题,提出一种基于自适应权重优化的多任务深度学习模型。该模型采用包含3种病害和1种健康状态的甘蔗叶片图像数据集,通过卷积神经网络(CNN)和多任务学习(MTL)实现病害识别。在模型训练过程中,为应对不同任务间的不平衡问题,引入了自适应权重优化方法。实验结果表明,该模型能显著提高甘蔗病害识别准确率,并在多任务之间实现平衡,为甘蔗智能化种植发展提供一定的借鉴。

This work proposes a multi-task deep learning model based on adaptive weight optimization,in order to solve the problem of the accuracy of sugarcane disease recognition and the balance between tasks in the agricultural field.The model uses a sugarcane leaf image dataset containing 3 diseases and 1 health state,and realizes disease recognition through convolutional neural network(CNN)and multi-task learning(MTL).In the process of model training,in order to deal with the imbalance between different tasks,an adaptive weight optimization method is introduced.The experimental results show that the model can significantly improve the accuracy of sugarcane disease identification,and achieve the balance between multiple tasks,providing a certain reference for the development of sugarcane intelligent planting.