Random Erasing Data Augmentation
By Zhun Zhong, Liang Zheng, Guoliang Kang, Shaozi Li, Yi Yang. 2017
4,240 citations recorded by the source corpus.
In this paper, we introduce Random Erasing, a new data augmentation method for training the convolutional neural network (CNN). In training, Random Erasing randomly selects a rectangle region in an image and erases its pixels with random values. In this process, training images with various levels of occlusion are generated, which reduces the risk of over-fitting and makes the model robust to occlusion. Random Erasing is parameter learning free, easy to implement, and can be integrated with most of the CNN-based recognition models. Albeit simple, Random Erasing is complementary to commonly used data augmentation techniques such as random cropping and flipping, and yields consistent improvement over strong baselines in image classification, object detection and person re-identification. Code is available at: https://github.com/zhunzhong07/Random-Erasing.
Subfields: Computer Vision and Pattern Recognition, Media Technology
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Computer Vision and Pattern Recognition
- Luc Van Gool
- Dacheng Tao
- Yi Yang
- Ming-Hsuan Yang
- Xiaogang Wang
- Trevor Darrell
- Alan L. Yuille
- Jitendra Malik
In Media Technology
- Jingdong Wang
- Andrew Zisserman
- Qi Tian
- Cordelia Schmid
- Herv'e J'egou
- Chunhua Shen
- Yi Yang
- Jitendra Malik
Citations in the corpus
- The Many Faces of Robustness: A Critical Analysis of Out-of-Distribution Generalization (2020, methodology): “We also evaluate with Random Erasure augmentation, which deletes rectangles within the image, to simulate occlusion [40].”
- Early Convolutions Help Transformers See Better (2021, background): “, removing erasing data augmentation [54] causes a catastrophic drop in ImageNet accuracy in [41].”
- VOLO: Vision Outlooker for Visual Recognition (2021, methodology): “For data augmentation methods, we use CutOut [74], RandAug [11], and the Token Labeling objective with MixToken [32].”
- Training data-efficient image transformers & distillation through attention (2020, methodology): “We also used random erasing [59] in the image domain.”
- Grafit: Learning fine-grained image representations with coarse labels (2020, positive): “The data augmentation consists of random resized crop, RandAugment [14] and Erasing [70].”
- Randaugment: Practical automated data augmentation with a reduced search space (2019, background): “dation accuracy [7, 49], robustness [40, 46, 9], or both [25].”
- CutMix: Regularization Strategy to Train Strong Classifiers With Localizable Features (2019, background): “While regional dropout strategies have shown improvements of classification and localization performances to a certain degree, deleted regions are usually zeroed-out [3, 32] or filled with random noise [50], greatly reducing the proportion of informative pixels on training images.”
- mixup: Beyond Empirical Risk Minimization (2017, methodology): “Following Zhang et al. (2017), we evaluate the robustness of ERM and mixup models against randomly corrupted labels.”
- Pyramid Vision Transformer: A Versatile Backbone for Dense Prediction without Convolutions (2021, methodology): “We follow DeiT [62] and apply random cropping, random horizontal flipping [58], label-smoothing regularization [59], mixup [76], CutMix [75], and random erasing [80] as data augmentations.”
- AugMix: A Simple Data Processing Method to Improve Robustness and Uncertainty (2019, methodology): “Random occlusion techniques such as Cutout can also improve accuracy on clean data (Devries & Taylor, 2017; Zhong et al., 2017).”
- Faster AutoAugment: Learning Augmentation Strategies using Backpropagation (2019, background): “[37,6] cut out a random patch from an image and replaced it with random noise or a constant value.”
- Greedy AutoAugment (2019, background): “Cutout[28], [22] Changing a random square patch of the image to gray pixels.”
- Searching for a Robust Neural Architecture in Four GPU Hours (2019, positive): “We use the standard preprocessing and data augmentation, i.e., randomly cropping, horizontally flipping, normalization, and CutOut [8, 43].”
- Style transfer-based image synthesis as an efficient regularization technique in deep learning (2019, background): “Another approach is a random erasing technique proposed in [19], fast and relatively easy to implement yet giving good results in generalization ability of CNN’s.”
- Learning data augmentation policies using augmented random search (2018, background): “(2) Sample crop and interpolation, such as random Eraser [9], Mixup [10].”
- DyTox: Transformers for Continual Learning with DYnamic TOken eXpansion (2021, methodology): “In DyTox, we applied the standard data augmentation of DeiT [25] but we removed the pixel erasing [32], MixUp [30], and Cut-Mix [29] augmentations for fair comparison.”