Category: 07. Data augmentation

Data augmentation is a very useful technique that can help to improve the translational invariance of convolutional neural networks (CNN). RandAugment is a stochastic data augmentation routine for vision data and was proposed in RandAugment: Practical automated data augmentation with a reduced search space. It is composed of strong augmentation transforms like color jitters, Gaussian blurs,…

Introduction mixup is a domain-agnostic data augmentation technique proposed in mixup: Beyond Empirical Risk Minimization by Zhang et al. It’s implemented with the following formulas: (Note that the lambda values are values with the [0, 1] range and are sampled from the Beta distribution.) The technique is quite systematically named. We are literally mixing up the features and their corresponding labels.…

Introduction CutMix is a data augmentation technique that addresses the issue of information loss and inefficiency present in regional dropout strategies. Instead of removing pixels and filling them with black or grey pixels or Gaussian noise, you replace the removed regions with a patch from another image, while the ground truth labels are mixed proportionally to…