Data Augmentation by Pairing Samples for Images Classification

By Hiroshi Inoue. 2018

470 citations recorded by the source corpus.

Data augmentation is a widely used technique in many machine learning tasks, such as image classification, to virtually enlarge the training dataset size and avoid overfitting. Traditional data augmentation techniques for image classification tasks create new samples from the original training data by, for example, flipping, distorting, adding a small amount of noise to, or cropping a patch from an original image. In this paper, we introduce a simple but surprisingly effective data augmentation technique for image classification tasks. With our technique, named SamplePairing, we synthesize a new sample from one image by overlaying another image randomly chosen from the training data (i.e., taking an average of two images for each pixel). By using two images randomly selected from the training set, we can generate N^2 new samples from N training samples. This simple data augmentation technique significantly improved classification accuracy for all the tested datasets; for example, the top-1 error rate was reduced from 33.5% to 29.0% for the ILSVRC 2012 dataset with GoogLeNet and from 8.22% to 6.93% in the CIFAR-10 dataset. We also show that our SamplePairing technique largely improved accuracy when the number of samples in the training set was very small. Therefore, our technique is more valuable for tasks with a limited amount of training data, such as medical imaging tasks.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition

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

Citations in the corpus

ScaleNet: An Unsupervised Representation Learning Method for Limited Information (2023, background): “Multiple studies employed data augmentation techniques such as scaling, rotating, cropping, and generating synthetic samples to produce more training samples [22,45,23].”
AutoAugment: Learning Augmentation Policies from Data (2018, methodology): “We additionally used two other promising augmentation techniques: Cutout [25] and SamplePairing [51].”
DADA: Differentiable Automatic Data Augmentation (2020, neutral): “Specifically, our search space contains the same 15 data augmentation operations as PBA [10], which is also the same as AA and Fast AA except that the SamplePairing [11] is removed.”
Adversarial AutoAugment (2019, methodology): “There are totally 16 image operations in our search space, including ShearX/Y, TranslateX/Y, Rotate, AutoContrast, Invert, Equalize, Solarize, Posterize, Contrast, Color, Brightness, Sharpness, Cutout (Devries & Taylor, 2017) and Sample Pairing (Inoue, 2018).”
Faster AutoAugment: Learning Augmentation Strategies using Backpropagation (2019, methodology): “and sample pairing [13] following [5,12,18].”
Greedy AutoAugment (2019, methodology): “Data augmentation refers to the practice of applying a series of standard transformations (wiredfool et al., 2016; DeVries and Taylor, 2017b; Inoue, 2018) to the given image data.”
Population Based Augmentation: Efficient Learning of Augmentation Policy Schedules (2019, methodology): “These include ShearX/Y, TranslateX/Y, Rotate, AutoContrast, Invert, Equalize, Solarize, Posterize, Contrast, Color, Brightness, Sharpness, Cutout (Devries & Taylor, 2017), and Sample-Pairing (Inoue, 2018).”
A survey on Image Data Augmentation for Deep Learning (2019, methodology): “The reduced CIFAR-10 results demonstrate the usefulness of the SamplePairing technique in limited data applications (Fig.”
Learning Data Augmentation Strategies for Object Detection (2019, background): “Data augmentation strategies for vision models are often focused on the image classification domain [21, 15, 41, 3, 6, 17, 28].”
On Mixup Training: Improved Calibration and Predictive Uncertainty for Deep Neural Networks (2019, background): “The experiments in [37] and related work in [14, 33, 10] show noticeable performance gains in various image classification tasks.”
Affinity and Diversity: Quantifying Mechanisms of Data Augmentation (2020, methodology): “Mixup (Zhang et al., 2017) and Sample Pairing (Inoue, 2018) are two augmentation methods that use convex combinations of training samples.”
A Comprehensive Survey of Few-shot Learning: Evolution, Applications, Challenges, and Opportunities (2022, methodology): “Random erasure [42] and random cropping [43], [53] are classical algorithms by simulating the images with different degrees of damage and thus improving the generalization of the model.”
Improved Mixed-Example Data Augmentation (2018, methodology): “Rather, the space of useful mixed-example data augmentation appears to be much larger than realized in previous work [27, 24, 12] — with the exception of “Rand.”
A Survey on Data Augmentation for Text Classification (2021, unknown): “If the amount of data is taken into consideration, it stands out that certain classification problems are often heavily unbalanced, for instance, only a small amount is relevant (positive) while the irrelevant (negative) data is prevalent [26].”
Learning to Recombine and Resample Data for Compositional Generalization (2020, methodology): “Existing proposals for sequence models are mainly rule-based—in sequence modeling problems, specifying a synchronous context-free grammar (Jia & Liang, 2016) or string rewriting system (Andreas, 2020) (Inoue, 2018) and machine translation (Fadaee et al., 2017).”
MPViT: Multi-Path Vision Transformer for Dense Prediction (2021, methodology): “Following the training recipe as in CoaT [65] and DeiT [14], we perform the same data augmentations such as MixUP [28], CutMix [70], random erasing [72], repeated augmentation [24], and label smoothing [48].”
MFC-GAN: Class-imbalanced dataset classification using Multiple Fake Class Generative Adversarial Network (2019, methodology): “P Data augmentation technique can be a source of more training data [6] or a regularizer [5] thereby improving generalization.”
Image Data Augmentation for Deep Learning: A Survey (2022, background): “In [Inoue, 2018], authors enlarge the dataset by synthesizing every new image with two images randomly selected in the training set, known as pairing samples.”