Fast AutoAugment

Data augmentation is an essential technique for improving generalization ability of deep learning models. Recently, AutoAugment has been proposed as an algorithm to automatically search for augmentation policies from a dataset and has significantly enhanced performances on many image recognition tasks. However, its search method requires thousands of GPU hours even for a relatively small dataset. In this paper, we propose an algorithm called Fast AutoAugment that finds effective augmentation policies via a more efficient search strategy based on density matching. In comparison to AutoAugment, the proposed algorithm speeds up the search time by orders of magnitude while achieves comparable performances on image recognition tasks with various models and datasets including CIFAR-10, CIFAR-100, SVHN, and ImageNet.

PDF Abstract NeurIPS 2019 PDF NeurIPS 2019 Abstract

Results from the Paper


Task Dataset Model Metric Name Metric Value Global Rank Uses Extra
Training Data
Result Benchmark
Image Classification CIFAR-10 PyramidNet+ShakeDrop (Fast AA) Percentage correct 98.3 # 46
Image Classification CIFAR-100 PyramidNet+ShakeDrop (Fast AA) Percentage correct 88.3 # 38
Image Classification ImageNet ResNet-50 (Fast AA) Top 1 Accuracy 77.6% # 852
Image Classification ImageNet ResNet-200 (Fast AA) Top 1 Accuracy 80.6% # 684
Data Augmentation ImageNet ResNet-50 (Fast AA) Accuracy (%) 77.6 # 12
Data Augmentation ImageNet ResNet-200 (Fast AA) Accuracy (%) 80.6 # 4
Image Classification SVHN Wide-ResNet-28-10 (Fast AA) Percentage error 1.1 # 2

Methods