AutoAugment: Learning Augmentation Strategies From Data

By Ekin Dogus Cubuk, Barret Zoph, Dandelion Mané, Vijay Vasudevan, Quoc V. Le. 2019

2,652 citations recorded by the source corpus.

Data augmentation is an effective technique for improving the accuracy of modern image classifiers. However, current data augmentation implementations are manually designed. In this paper, we describe a simple procedure called AutoAugment to automatically search for improved data augmentation policies. In our implementation, we have designed a search space where a policy consists of many sub-policies, one of which is randomly chosen for each image in each mini-batch. A sub-policy consists of two operations, each operation being an image processing function such as translation, rotation, or shearing, and the probabilities and magnitudes with which the functions are applied. We use a search algorithm to find the best policy such that the neural network yields the highest validation accuracy on a target dataset. Our method achieves state-of-the-art accuracy on CIFAR-10, CIFAR-100, SVHN, and ImageNet (without additional data). On ImageNet, we attain a Top-1 accuracy of 83.5% which is 0.4% better than the previous record of 83.1%. On CIFAR-10, we achieve an error rate of 1.5%, which is 0.6% better than the previous state-of-the-art. Augmentation policies we find are transferable between datasets. The policy learned on ImageNet transfers well to achieve significant improvements on other datasets, such as Oxford Flowers, Caltech-101, Oxford-IIT Pets, FGVC Aircraft, and Stanford Cars.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Representation & Self-Supervised Learning

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Computer Vision and Pattern Recognition

In Representation & Self-Supervised Learning

Citations in the corpus

A Simple Framework for Contrastive Learning of Visual Representations (2020, neutral): “We use a simpler data augmentation policy, while they use FastAutoAugment for their best result.”
Shortcut learning in deep neural networks (2020, background): “iance [123] can be applied. This strategy can be applied to almost any well-understood transformation of the inputs and ﬁnds its probably most general form in auto-augment as an augmentation strategy [124]. Extreme dataaugmentation strategies are also the core ingredient of the most successful semi-”
On Robustness and Transferability of Convolutional Neural Networks (2020, methodology): “We evaluate ResNet-50 [24] and six EfficientNet (B0 through B5) models [60] including variants using AutoAugment [10] and AdvProp [69], which have been trained on IMAGENET.”
Visual Transformers: Token-based Image Representation and Processing for Computer Vision (2020, methodology): “We use label smoothing and AutoAugment [36] in the training.”
Self-Training With Noisy Student Improves ImageNet Classification (2019, neutral): “2% AmoebaNet-C [17] 155M 83.”
Big Self-Supervised Models are Strong Semi-Supervised Learners (2020, methodology): “For supervised models on 1%/10% labels, AutoAugment [35] and label smoothing [36] are used.”
FixMatch: Simplifying Semi-Supervised Learning with Consistency and Confidence (2020, methodology): “To do so, a wide range of transformation magnitude values is divided into bins (as in AutoAugment [9]) and a weight (initially set to 1 ) is assigned to each bin.”
FILIP: Fine-grained Interactive Language-Image Pre-Training (2021, methodology): “We apply AutoAugment (Krizhevsky et al., 2012; Sato et al., 2015; Cubuk et al., 2019; Hoffer et al., 2020) for image augmentation, following the SOTA vision recognition methods (Touvron et al., 2021; Xie et al., 2020b).”
CrossTransformers: spatially-aware few-shot transfer (2020, methodology): “Therefore, for two experiments, we employ augmentation using the settings discovered in BOHB [68] (via Auto-Augment [19] on the validation set), with an extra stage that randomly downsamples and then upsamples images, which we find helpful as our network operates at higher resolution than many of th”
Dynamic Neural Networks: A Survey (2021, background): “Dynamic networks are compatible with most advanced techniques in deep learning, including architecture design [4], [5], optimization algorithms [23], [24] and data preprocessing [25], [26], which ensures that they can benefit from the most recent advances in the field to achieve state-of-the-art per”
AutoML: A Survey of the State-of-the-Art (2019, background): “[45] propose Autoaugment to automatically search for improved data-augmentation policies.”
Toward Causal Representation Learning (2021, methodology): “Instead of computing the maximum over all distributions in E, one can relax the problem by sampling from the interventional distributions and optimize an expectation over the different augmented images on a suitably chosen subset [39], using a search algorithm, such as RL [49] or an algorithm based ”
FBNetV3: Joint Architecture-Recipe Search using Neural Acquisition Function (2020, methodology): “These methods use policy search [4], population-based training [12], and Bayesian-based augmentation [35].”
A Comprehensive Survey of Neural Architecture Search (2020, background): “emantic segmentation [63, 64, 120], adversarial learning [53], architectural scaling [114, 122, 124], multi-objective optimization [39, 115, 125], platform-aware [28, 34, 103, 117], data augmentation [121, 123] and so on. In addition, there is some work to consider how to strike a balance between pe”
UniformAugment: A Search-free Probabilistic Data Augmentation Approach (2020, methodology): “Recent advancements in such search-based methods, inspired by Neural Architecture Search (NAS) [20], gave rise to algorithms like AutoAugment (AA) [3] which uses Reinforcement Learning (RL) to search for an optimal augmentation policy for a given task.”
BigNAS: Scaling Up Neural Architecture Search with Big Single-Stage Models (2020, background): “4 epochs; swish activation [26] and AutoAugment policy [6].”
DADA: Differentiable Automatic Data Augmentation (2020, neutral): “Recently, inspired by the neural architecture search (NAS) algorithms, some methods [3, 10, 15, 16] attempted to automate learning data augmentation policies.”
Fast AutoAugment (2019, methodology): “Dataset AutoAug [5] Fast AutoAug CIFAR-10 5000 3.”
Towards Causal Representation Learning (2021, methodology): “Instead of computing the maximum over all distributions in E, one can relax the problem by sampling from the interventional distributions and optimize an expectation over the different augmented images on a suitably chosen subset [38], using a search algorithm like reinforcement learning [48] or an”
Neural Prototype Trees for Interpretable Fine-grained Image Recognition (2020, background): “[3] David Alvarez-Melis and Tommi S Jaakkola.”