See Better Before Looking Closer: Weakly Supervised Data Augmentation Network for Fine-Grained Visual Classification

Data augmentation is usually adopted to increase the amount of training data, overfitting and improve the performance of deep models. However, in, random data augmentation, such as random image cropping, is-efficiency and might introduce many uncontrolled background noises. In this, we propose Weakly Supervised Data Augmentation Network (WS-DAN) to the potential of data augmentation. Specifically, for each training, we first generate attention maps to represent the object's parts by weakly supervised learning. Next, we augment the image by these attention maps, including attention cropping and attention. The proposed WS-DAN improves the classification accuracy in two. In the first stage, images can be seen better since more discriminative' features will be extracted. In the second stage, attention regions accurate location of object, which ensures our model to look at the closer and further improve the performance. Comprehensive experiments in fine-grained visual classification datasets show that our WS-DAN the state-of-the-art methods, which demonstrates its effectiveness.