Open AccessEfficientNet based recognition of maize diseases by leaf image classification
Author(s) -
Jiangchuan Liu,
Mantao Wang,
Bao Lie,
Xiaofan Li
Publication year - 2020
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/1693/1/012148
Subject(s) - softmax function , computer science , classifier (uml) , artificial intelligence , robustness (evolution) , pattern recognition (psychology) , transfer of learning , training set , test data , machine learning , deep learning , biochemistry , chemistry , gene , programming language
For the research on the recognition and classification of maize leaf disease pictures, this paper proposes a method of fine-tuning model parameters based on transfer learning EfficientNet, which can improve the accuracy and speed of network recognition for a small sample of maize disease dataset. First of all, perform data cleaning and data augmentation on the dataset to obtain richer image data; then, transfer the pre-trained model obtained by EfficientNet training on ImageNet to this model method; finally, the last layer of EfficientNet classifier replace with 8 classes of softmax classifier, and train the entire network to obtain a training model for maize disease prediction. In order to verify the robustness and accuracy of the method proposed in this paper, test verification was carried out in the test dataset with VGG-16, Inception-v3 and Resnet-50, respectively. The experimental results show that the training speed of the network model proposed in this paper has been significantly improved, and its recognition accuracy is far better than other networks with a maximum of 98.52%, which can realize agricultural production applications.