Adversarial Learning of General Transformations for Data Augmentation
This addresses data augmentation for image classification, but it is incremental as it builds on existing methods like spatial transformer networks.
The paper tackles the problem of overfitting in convolutional neural networks by learning data augmentation transformations directly from training data, using an encoder-decoder and spatial transformer network, and shows it outperforms generative methods and matches predefined transformations in image classification.
Data augmentation (DA) is fundamental against overfitting in large convolutional neural networks, especially with a limited training dataset. In images, DA is usually based on heuristic transformations, like geometric or color transformations. Instead of using predefined transformations, our work learns data augmentation directly from the training data by learning to transform images with an encoder-decoder architecture combined with a spatial transformer network. The transformed images still belong to the same class but are new, more complex samples for the classifier. Our experiments show that our approach is better than previous generative data augmentation methods, and comparable to predefined transformation methods when training an image classifier.