CutMix: Regularization Strategy to Train Strong Classifiers With Localizable Features

Abstract

Regional dropout strategies have been proposed to enhance the performance of convolutional neural network classifiers. They have proved to be effective for guiding the model to attend on less discriminative parts of objects (e.g. leg as opposed to head of a person), thereby letting the network generalize better and have better object localization capabilities. On the other hand, current methods for regional dropout removes informative pixels on training images by overlaying a patch of either black pixels or random noise. Such removal is not desirable because it leads to information loss and inefficiency during training. We therefore propose the CutMix augmentation strategy: patches are cut and pasted among training images where the ground truth labels are also mixed proportionally to the area of the patches. By making efficient use of training pixels and retaining the regularization effect of regional dropout, CutMix consistently outperforms the stateof-the-art augmentation strategies on CIFAR and ImageNet classification tasks, as well as on the ImageNet weaklysupervised localization task. Moreover, unlike previous augmentation methods, our CutMix-trained ImageNet classifier, when used as a pretrained model, results in consistent performance gains in Pascal detection and MS-COCO image captioning benchmarks. We also show that CutMix improves the model robustness against input corruptions and its out-of-distribution detection performances.

Attentive Cutmix: An Enhanced Data Augmentation Approach for Deep Learning Based Image Classification

“…CutMix Figure 1: Comparison of our proposed Attentive CutMix with Mixup [5], Cutout [1] and CutMix [3].…”

“…Here, x ∈ R W ×H×C is the training image and y is the training label. Similar to CutMix [3], we define this combining operation as,…”

ROIMIX: Proposal-Fusion Among Multiple Images for Underwater Object Detection

“…This helps the model attend to the most discriminative part of the objects, but it can result in the loss of information. Moreover, an advanced version CutMix [11] cuts and pastes the patches among the training data, which greatly improves the model robustness against input corruption. For object detection, the detector adopts multiple augmentation strategies, such as photo metric distortion [23], image mirror [24] and multi-scale training [25].…”

“…In the image classification domain, translating and flipping are commonly used strategy to increase the amount of training data. Some works such as Mixup [10], CutMix [11] are devoted to creating better training data. We investigate the effect of deploying data augmentation in training object detectors.…”

Self-Supervised Contextual Data Augmentation for Natural Language Processing

“…In the CV field, image data can be easily augmented using various methods, such as cropping, rotation, scaling, shifting, and noise addition. Beyond the technique of augmenting the data similar to the original image, methods for augmenting the image with a large difference from the original image are being studied, as in Cutout [25], Mixup [26], and CutMix [27].…”