基于迁移学习的番茄叶片病害图像分类

Image classification of tomato leaf diseases based on transfer learning

针对卷积神经网络对番茄病害识别需训练参数较多,训练非常耗时的问题,将迁移学习应用于AlexNet卷积神经网络,对病害叶片和健康叶片共10种类别的番茄叶片进行分类研究。使用14 529张番茄叶片病害图像,随机选择70%作为训练集,30%作为验证集,对AlexNet卷积神经网络模型结构进行迁移,利用在Imagenet图像数据集上训练成熟的AlexNet模型和其参数对番茄叶片病害识别。在训练过程中,固定低层网络参数不变,微调高层网络参数,将番茄病害图像输入到网络中训练网络高层参数,用训练好的模型对10种类别的番茄叶片分类,并进行了20组试验。结果表明:该算法在训练迭代474次时使网络模型很好的收敛,网络对验证集的测试平均准确率达到95.62%,与从零开始训练的AlexNet卷积神经网络相比,本研究算法缩短了训练时间,平均准确率提高了5.6%。采用迁移学习所建立的病害分类模型能够对10种类别的番茄叶片病害快速准确地分类。

In view of the convolutional neural network for identifying tomato diseases needs more training parameters and the training is very time-consuming,transfer learning is applied to the AlexNet convolutional neural network,and 9 types of disease leaves including healthy tomato leaves are classified.Using 14 529 tomato leaf images,70% randomly selected as the training set and 30% as the validation set,the AlexNet convolutional neural network model structure was migrated and the mature AlexNet model and its parameters were trained on the Imagenet image dataset for tomato disease identification.During the training process,the fixed low-level network parameters are unchanged,the high-level network parameters are fine-tuned,the tomato disease image is input into the network to train the high-level parameters of the network,and the trained model is used to classify the 10 types of tomato leaves,and 20 groups of experiments are carried out.The results show that the proposed algorithm converges well in the network of 474 training iterations,and the average accuracy of the network to the verification set is 95.62%.Compared with the AlexNet convolutional neural network trained from zero,the training time is shortened and the average accuracy is increased by 5.6%.In conclusion,the disease classification model established by transfer learning can quickly and accurately classify 10 types of tomato leaf diseases.