The Performance Research of the Data Augmentation Method for Image Classification

Figure 3

The pipeline of our experiments. Collected images from Pest, ImageNet, and VOC, respectively. Augmentation methods—color, contrast, flip, shift, and rotation only used once each time to train the network models AlexNet, VGG16, and InceptionV3.