Ecosyste.ms: Awesome

An open API service indexing awesome lists of open source software.

Awesome Lists | Featured Topics | Projects

https://github.com/JasonZhang156/awesome-mixed-sample-data-augmentation

A collection of awesome things about mixed sample data augmentation
https://github.com/JasonZhang156/awesome-mixed-sample-data-augmentation

List: awesome-mixed-sample-data-augmentation

data-augmentation mixed-sample

Last synced: 3 months ago
JSON representation

A collection of awesome things about mixed sample data augmentation

Awesome Lists containing this project

README 

        
# awesome-mixed-sample-data-augmentation



This repo is a collection of awesome things about mixed sample data augmentation, including papers, code, etc.



_ _ _



## Basic Method

We introduce a basic usage of mixed sample data augmentation, which was first proposed in **mixup: Beyond Empirical Risk Minimization [[ICLR2018]](https://arxiv.org/abs/1710.09412) [[code]](https://github.com/facebookresearch/mixup-cifar10)**.



### Formulation



In mixup, the virtual training feature-target samples are produced as,



```

x˜ = λxi + (1 − λ)xj

y˜ = λyi + (1 − λ)yj

```



where (xi, yi) and (xj, yj) are two feature-target samples drawn at random from the training data, λ∈[0, 1]. The mixup hyper-parameter α controls the strength of interpolation between feature-target pairs and λ∼Beta(α, α).



### Training Pipeline



The simple and basic training pipeline is shown as the following Figure,



![](/image/mixup_pipeline.png)



### Core Code



The few lines of code necessary to implement mixup training in PyTorch



```Python

for (x1, y1), (x2, y2) in zip(loader1, loader2): 

  lam = numpy.random.beta(alpha, alpha) 

  x = Variable(lam * x1 + (1. - lam) * x2) 

  y = Variable(lam * y1 + (1. - lam) * y2) 

  optimizer.zero_grad() 

  loss(net(x), y).backward()

  optimizer.step()

```



_ _ _



## Application

### Classification Tasks (Image, Text, Audio, ...)

- AugMix: A Simple Data Processing Method to Improve Robustness and Uncertainty [[ICLR2020]](https://arxiv.org/pdf/1912.02781.pdf) [[code]](https://github.com/google-research/augmix)

- SuperMix: Supervising the Mixing Data Augmentation [[Arxiv2020]](https://arxiv.org/pdf/2003.05034.pdf) [[code]](https://github.com/alldbi/SuperMix)

- Nonlinear Mixup: Out-Of-Manifold Data Augmentation for Text Classification [[AAAI2020]](https://www.aaai.org/Papers/AAAI/2020GB/AAAI-GuoH.6040.pdf)

- Adversarial Vertex Mixup: Toward Better Adversarially Robust Generalization [[Arxiv2020]](https://arxiv.org/pdf/2003.02484.pdf)

- Attribute Mix: Semantic Data Augmentation for Fine-grained Recognition [[Arxiv2020]](https://arxiv.org/pdf/2004.02684.pdf)

- Understanding and Enhancing Mixed Sample Data Augmentation [[Arxiv2020]](https://arxiv.org/abs/2002.12047) [[code]](https://github.com/ecs-vlc/FMix)

- Attentive CutMix: An Enhanced Data Augmentation Approach for Deep Learning Based Image Classification [[ICASSP2020]](https://arxiv.org/abs/2003.13048)

- Mixup-breakdown: a consistency training method for improving generalization of speech separation models [[ICASSP2020]](https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=9054719)

- Cutmix: Regularization strategy to train strong classifiers with localizable features [[ICCV2019]](http://openaccess.thecvf.com/content_ICCV_2019/papers/Yun_CutMix_Regularization_Strategy_to_Train_Strong_Classifiers_With_Localizable_Features_ICCV_2019_paper.pdf) [[code]](https://github.com/clovaai/CutMix-PyTorch)

- Improved Mixed-Example Data Augmentation [[WACV2019]](https://arxiv.org/abs/1805.11272) [[code]](https://github.com/ceciliaresearch/MixedExample)

- Patch-level Neighborhood Interpolation: A General and Effective Graph-based Regularization Strategy [[Arxiv2019]](https://arxiv.org/pdf/1911.09307.pdf)

- Target-Directed MixUp for Labeling Tangut Characters [[ICDAR2019]](https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=8978040)

- Manifold Mixup improves text recognition with CTC loss [[Arixv2019]](https://arxiv.org/pdf/1903.04246.pdf)

- Manifold Mixup: Better Representations by Interpolating Hidden States [[ICML2019]](https://arxiv.org/abs/1806.05236) [[code]](https://github.com/vikasverma1077/manifold_mixup)

- Data augmentation using random image cropping and patching for deep CNNs [[TCSVT2019]](https://arxiv.org/abs/1811.09030) [[code]](https://github.com/jackryo/ricap)

- MixUp as Locally Linear Out-Of-Manifold Regularization [[AAAI2019]](https://www.aaai.org/ojs/index.php/AAAI/article/download/4256/4134)

- On Adversarial Mixup Resynthesis [[NeurIPS2019]](http://papers.nips.cc/paper/8686-on-adversarial-mixup-resynthesis.pdf) [[code]](https://github.com/christopher-beckham/amr)

- On mixup training: Improved calibration and predictive uncertainty for deep neural networks [[NeurIPS2019]](http://papers.nips.cc/paper/9540-on-mixup-training-improved-calibration-and-predictive-uncertainty-for-deep-neural-networks.pdf)

- mixup: Beyond Empirical Risk Minimization [[ICLR2018]](https://arxiv.org/abs/1710.09412) [[code]](https://github.com/facebookresearch/mixup-cifar10)

- Learning from between-class examples for deep sound recognition [[ICLR2018]](https://arxiv.org/abs/1711.10282) [[code]](https://github.com/mil-tokyo/bc_learning_sound/)

- Between-class Learning for Image Classification [[CVPR2018]](http://openaccess.thecvf.com/content_cvpr_2018/papers/Tokozume_Between-Class_Learning_for_CVPR_2018_paper.pdf) [[code]](https://github.com/mil-tokyo/bc_learning_image/)

- Data Augmentation by Pairing Samples for Images Classification [[Arxiv2018]](https://arxiv.org/abs/1801.02929)

- Rare Sound Event Detection Using Deep Learning and Data Augmentation [[Interspeech2019]](https://www.isca-speech.org/archive/Interspeech_2019/pdfs/1985.pdf)

- Mixup Learning Strategies for Text-independent Speaker Verification [[Interspeech2019]](https://pdfs.semanticscholar.org/0bc3/f8c6bc1f3568aac96d3ad0632ebe41134611.pdf)

- Acoustic Scene Classification with Mismatched Devices Using CliqueNets and

Mixup Data Augmentation [[Interspeech2019]](https://www.isca-speech.org/archive/Interspeech_2019/pdfs/3002.pdf)

- Deep Convolutional Neural Network with Mixup for Environmental Sound Classification [[PRCV2018]](https://arxiv.org/abs/1808.08405)

- Speaker Adaptive Training and Mixup Regularization for Neural Network Acoustic Models in Automatic Speech Recognition [[Interspeech2018]](https://www.isca-speech.org/archive/Interspeech_2018/pdfs/2209.pdf)

- An investigation of mixup training strategies for acoustic models in ASR [[Interspeech2018]](https://www.researchgate.net/profile/Ivan_Medennikov/publication/327389098_An_Investigation_of_Mixup_Training_Strategies_for_Acoustic_Models_in_ASR/links/5bc86248a6fdcc03c78f5a44/An-Investigation-of-Mixup-Training-Strategies-for-Acoustic-Models-in-ASR.pdf) [[code]](https://github.com/speechpro/mixup)

- Understanding Mixup Training Methods [[IEEE ACCESS 2018]](https://ieeexplore.ieee.org/stamp/stamp.jsp?arnumber=8478159)



### Semi-Supervised Learning

- Rethinking Image Mixture for Unsupervised Visual Representation Learning [[Arxiv2020]](https://arxiv.org/pdf/2003.05438.pdf)

- FocalMix: Semi-Supervised Learning for 3D Medical Image Detection [[Arxiv2020]](https://arxiv.org/pdf/2003.09108.pdf)

- ReMixMatch: Semi-Supervised Learning with Distribution Alignment and Augmentation Anchoring [[ICLR2020]](https://arxiv.org/pdf/1911.09785.pdf) [[code]](https://github.com/google-research/remixmatch)

- DivideMix: Learning with Noisy Labels as Semi-supervised Learning [[ICLR2020]](https://arxiv.org/pdf/2002.07394.pdf) [[code]](https://github.com/LiJunnan1992/DivideMix)

- OpenMix: Reviving Known Knowledge for Discovering Novel Visual Categories in An Open World [[Arxiv2020]](https://arxiv.org/pdf/2004.05551.pdf)

- MixPUL: Consistency-based Augmentation for Positive and Unlabeled Learning [[Arxiv2020]](https://arxiv.org/pdf/2004.09388.pdf)

- ROAM: Random Layer Mixup for Semi-Supervised Learning in Medical Imaging [[Arxiv2020]](https://arxiv.org/pdf/2003.09439.pdf)

- Interpolation Consistency Training for Semi-Supervised Learning [[IJCAI2019]](https://arxiv.org/abs/1903.03825) [[code]](https://github.com/vikasverma1077/ICT)

- RealMix: Towards Realistic Semi-Supervised Deep Learning Algorithms [[Arxiv2019]](https://arxiv.org/pdf/1912.08766.pdf) [[code]](https://github.com/uizard-technologies/realmix)

- Unifying semi-supervised and robust learning by mixup [[ICLR Workshop 2019]](https://openreview.net/pdf?id=r1gp1jRN_4)

- On Adversarial Mixup Resynthesis [[NeurIPS2019]](http://papers.nips.cc/paper/8686-on-adversarial-mixup-resynthesis.pdf) [[code]](https://github.com/christopher-beckham/amr)

- Unifying semi-supervised and robust learning by mixup [[ICLR2019 Workshop]](https://openreview.net/pdf?id=r1gp1jRN_4)

- Mixmatch: A holistic approach to semi-supervised learning [[NeurIPS2019]](https://papers.nips.cc/paper/8749-mixmatch-a-holistic-approach-to-semi-supervised-learning.pdf) [[code]](https://github.com/google-research/mixmatch)

- Semi-Supervised and Task-Driven Data Augmentation [[IPMI2019]](https://arxiv.org/abs/1902.05396)



### Object Detection and Localization

- Mixup Regularization for Region Proposal based Object Detectors [[Arxiv2020]](https://arxiv.org/pdf/2003.02065.pdf)

- FocalMix: Semi-Supervised Learning for 3D Medical Image Detection [[Arxiv2020]](https://arxiv.org/pdf/2003.09108.pdf)

- Cutmix: Regularization strategy to train strong classifiers with localizable features [[ICCV2019]](http://openaccess.thecvf.com/content_ICCV_2019/papers/Yun_CutMix_Regularization_Strategy_to_Train_Strong_Classifiers_With_Localizable_Features_ICCV_2019_paper.pdf) [[code]](https://github.com/clovaai/CutMix-PyTorch)



### Natural Language Processing

- On mixup training: Improved calibration and predictive uncertainty for deep neural networks [[NeurIPS2019]](http://papers.nips.cc/paper/9540-on-mixup-training-improved-calibration-and-predictive-uncertainty-for-deep-neural-networks.pdf)



### Image Segmentation

- ROAM: Random Layer Mixup for Semi-Supervised Learning in Medical Imaging [[Arxiv2020]](https://arxiv.org/pdf/2003.09439.pdf)

- Improving Robustness of Deep Learning Based Knee MRI Segmentation: Mixup and Adversarial Domain Adaptation [[Arxiv2019]](https://arxiv.org/abs/1908.04126)

- Improving Data Augmentation for Medical Image Segmentation [[MIDL2018]](https://openreview.net/pdf?id=rkBBChjiG)



### Super Resolution

- Rethinking Data Augmentation for Image Super-resolution: A Comprehensive Analysis and a New Strategy [[Arxiv2020]](https://arxiv.org/pdf/2004.00448.pdf)



### Novelty Detection

- Multi-class Novelty Detection Using Mix-up Technique [[WACV2020]](http://openaccess.thecvf.com/content_WACV_2020/papers/Bhattacharjee_Multi-class_Novelty_Detection_Using_Mix-up_Technique_WACV_2020_paper.pdf) 



### Generative Adversarial Networks

- A U-Net Based Discriminator for Generative Adversarial Networks [[CVPR2020]](https://arxiv.org/abs/2002.12655)

- Mixed batches and symmetric discriminators for GAN training [[ICML2018]](https://arxiv.org/abs/1806.07185)

- mixup: Beyond Empirical Risk Minimization [[ICLR2018]](https://arxiv.org/abs/1710.09412) [[code]](https://github.com/facebookresearch/mixup-cifar10)



### Domain Adaptation

- Improve Unsupervised Domain Adaptation with Mixup Training [[Arxiv2020]](https://arxiv.org/abs/2001.00677)

- Adversarial Domain Adaptation with Domain Mixup [[AAAI2020]](https://arxiv.org/abs/1912.01805)



### Few-shot Learning

- Charting the Right Manifold: Manifold Mixup for Few-shot Learning [[WACV2020]](http://openaccess.thecvf.com/content_WACV_2020/papers/Mangla_Charting_the_Right_Manifold_Manifold_Mixup_for_Few-shot_Learning_WACV_2020_paper.pdf) [[code]](https://github.com/nupurkmr9/S2M2_fewshot)



### Machine Learning

- An Experimental Evaluation of Mixup Regression Forests [[EXPERT SYST APPL 2020]](https://lucykuncheva.co.uk/papers/jrmjaalkeswa20.pdf)



### Analysis

- Data Augmentation Revisited: Rethinking the Distribution Gap between Clean and Augmented Data [[Arxiv2019]](https://arxiv.org/pdf/1909.09148.pdf)