Authors: Changxu Cheng, Qiuhui Huang, Wuheng Xu and Hao Wang at Huazhong University of Science and Technology

Description: We use a simple VGG16 followed by Global Average Pooling and a linear classifier. Data augmentation, including random crop, changing hsv and perspective transformation, is imposed on the training data. Besides, we adopt grouping resizing strategy to deal with the trainging images. Specifically, we resize the images to a certain size according to their aspect ratios. In this way, only images with similar aspect ratios has the same final size.