RandAugment: Practical data augmentation with no separate search

By Ekin Dogus Cubuk, Barret Zoph, Jonathon Shlens, Quoc V. Le. 2019

265 citations recorded by the source corpus.

Subfields: 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 Representation & Self-Supervised Learning

Citations in the corpus

Unsupervised Data Augmentation for Consistency Training (2019, methodology): “We use a data augmentation method called RandAugment [15], which is inspired by AutoAugment [11].”
Contrastive Multiview Coding (2019, positive): “On top of it, strengthening data augmentation with RandAugment [13] yields better or comparable results to other state-of-the-art methods [40,79,87,26,48,18,28,4].”
A Survey of Data Augmentation Approaches for NLP (2021, background): “Whereas there are well-accepted frameworks for DA for CV (e.g. default augmentation libraries in PyTorch, RandAugment (Cubuk et al., 2020)), there are no such "generalized" DA techniques for NLP.”
What makes for good views for contrastive learning (2020, background): “, RA [11]) which results in smaller I(v 1 ;v cpc 2 ).”
Adversarial Filters of Dataset Biases (2020, methodology): “The EfﬁcientNet model is learned from scratch on a ﬁxed 20% sample of the Ima-geNet training set, using RandAugment data augmentation (Cubuk et al., 2019).”
Rainbow Memory: Continual Learning with a Memory of Diverse Samples (2021, methodology): “Along with the label mixing augmentation, we exploit the effects of composition of multiple data augmentations for enhancing the diversity, benefiting from conventional methods such as flipping, shearing, or color jittering and recent automated data augmentation researches [9, 10, 32].”
Supervised Contrastive Learning (2020, positive): “In light of the ﬁndings of [6] that self-supervised contrastive loss requires signiﬁcantly different data augmentation than cross-entropy loss, for the second stage we evaluate three different options: – AutoAugment: [9] – RandAugment: [10] – SimAugment: A variant of the strategy of [6] to sequentia”
Scaling Local Self-Attention for Parameter Efficient Visual Backbones (2021, methodology): “When label smoothing and RandAugment are added, HaloNet improves by 1.3% while ResNet improves by 0.8%.”
Adversarial Examples Improve Image Recognition (2019, background): “Recent works also demonstrate that it is possible to learn data augmentation policies automatically for achieving better performance on image classification [3, 4, 19, 21, 52] and object detection [53, 4].”
Supervision Accelerates Pre-training in Contrastive Semi-Supervised Learning of Visual Representations. (2020, background): “An additional key feature of FixMatch is the use of learned data augmentations [3, 10].”
Milking CowMask for Semi-Supervised Image Classification (2020, methodology): “AutoAugment (Cubuk et al., 2019a) and RandAugment (Cubuk et al., 2019b) are rich augmentation schemes that combine a number of image operations provided by the Pillow library (Lundh et al.).”
An Overview of Deep Semi-Supervised Learning (2020, methodology): “Unsupervised Data Augmentation [167] uses advanced data augmentation methods, such as AutoAugment [27], RandAugment [28] and Back Translation [41, 137] as perturbations for consistency training based SSL.”
Federated Semi-Supervised Learning with Inter-Client Consistency (2020, background): “…learned σ G as well as the aggregated ψ and top-H reliable ψ h 1: H to the next available clients ( H = 2 ). follows: where π ( u ) performs RandAugment (Cubuk et al., 2019) on unlabeld instance u , and ˆ y is the agreement-based pseudo label, where 1 ( · ) produces one-hot labels with given…”
i-Mix: A Domain-Agnostic Strategy for Contrastive Representation Learning (2021, unknown): “…must be understood as the original data, data augmentations are carefully designed using the domain knowledge about images (DeVries & Taylor, 2017b; Cubuk et al., 2019a;b; Zhong et al., 2020), speech (Amodei et al., 2016; Park et al., 2019), or natural languages (Zhang et al., 2015; Wei & Zou,…”