Improved Image Augmentation for Convolutional Neural Networks by Copyout and CopyPairing

Image augmentationis a widely used technique to improve the performance of convolutional neural networks (CNNs). In common image shifting, cropping, flipping, shearing and rotating are used for augmentation. But there are more advanced techniques like Cutout and SamplePairing. In this work we present two improvements of the state-of-the-art Cutout and SamplePairing techniques. Our new method called Copyout takes a square patch of another random training image and copies it onto a random location of each image used for training. The second technique we discovered is called CopyPairing. It combines Copyout and SamplePairing for further augmentationand even better performance. We apply different experiments with these augmentationtechniques on the CIFAR-10 dataset to evaluate and comparethem under different configurations. In our experiments we show that Copyout reduces the test error rate by 8.18% comparedwith Cutout and 4.27% comparedwith SamplePairing. CopyPairing reduces the test error rate by 11.97% comparedwith Cutout and 8.21% comparedwith SamplePairing. Copyout and CopyPairing implementations are available at https://github.com/t-systems-on-site-services-gmbh/coocop.