Does Data Augmentation Benefit from Split BatchNorms
arXiv:2010.07810
Abstract
Data augmentation has emerged as a powerful technique for improving the performance of deep neural networks and led to state-of-the-art results in computer vision. However, state-of-the-art data augmentation strongly distorts training images, leading to a disparity between examples seen during training and inference. In this work, we explore a recently proposed training paradigm in order to correct for this disparity: using an auxiliary BatchNorm for the potentially out-of-distribution, strongly augmented images. Our experiments then focus on how to define the BatchNorm parameters that are used at evaluation. To eliminate the train-test disparity, we experiment with using the batch statistics defined by clean training images only, yet surprisingly find that this does not yield improvements in model performance. Instead, we investigate using BatchNorm parameters defined by weak augmentations and find that this method significantly improves the performance of common image classification benchmarks such as CIFAR-10, CIFAR-100, and ImageNet. We then explore a fundamental trade-off between accuracy and robustness coming from using different BatchNorm parameters, providing greater insight into the benefits of data augmentation on model performance.
arXiv comment9 pages (+ 3 for references)
References in corpus (18)
- EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks
- mixup: Beyond Empirical Risk Minimization
- Improved Regularization of Convolutional Neural Networks with Cutout
- FixMatch: Simplifying Semi-Supervised Learning with Consistency and Confidence
- Wide Residual Networks
- Unsupervised Data Augmentation for Consistency Training
- Random Erasing Data Augmentation
- AugMix: A Simple Data Processing Method to Improve Robustness and Uncertainty
- Rethinking Pre-training and Self-training
- Revisiting Batch Normalization For Practical Domain Adaptation
- Supervised Contrastive Learning
- APAC: Augmented PAttern Classification with Neural Networks
- Improving Robustness Without Sacrificing Accuracy with Patch Gaussian Augmentation
- Affinity and Diversity: Quantifying Mechanisms of Data Augmentation
- Augment your batch: better training with larger batches
- Data Augmentation Revisited: Rethinking the Distribution Gap between Clean and Augmented Data
- Improving 3D Object Detection through Progressive Population Based Augmentation
- Split Batch Normalization: Improving Semi-Supervised Learning under Domain Shift