KeepAugment: A Simple Information-Preserving Data Augmentation Approach

2020/11/23 by Chengyue Gong, Dilin Wang, Gong, Chengyue +7 · 6 citations
Computer Science · #Advanced Neural Network Applications #Computer Vision and Pattern Recognition (cs.CV) #Domain Adaptation and Few-Shot Learning #FOS: Computer and information sciences #Visual Attention and Saliency Detection

paper · pdf · doi:10.48550/arxiv.2011.11778

openalex publication_date 2020/11/23 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28

Abstract

Data augmentation (DA) is an essential technique for training state-of-the-art deep learning systems. In this paper, we empirically show data augmentation might introduce noisy augmented examples and consequently hurt the performance on unaugmented data during inference. To alleviate this issue, we propose a simple yet highly effective approach, dubbed KeepAugment, to increase augmented images fidelity. The idea is first to use the saliency map to detect important regions on the original images and then preserve these informative regions during augmentation. This information-preserving strategy allows us to generate more faithful training examples. Empirically, we demonstrate our method significantly improves on a number of prior art data augmentation schemes, e.g. AutoAugment, Cutout, random erasing, achieving promising results on image classification, semi-supervised image classification, multi-view multi-camera tracking and object detection.

Citations

• Faster R-CNN: Towards Real-Time Object Detection with Region Proposal\n Networks
• Random Erasing Data Augmentation
• DropBlock: A regularization method for convolutional networks
• Unsupervised Data Augmentation for Consistency Training
• Deep Inside Convolutional Networks: Visualising Image Classification Models and Saliency Maps
• Adversarial AutoAugment
• GridMask Data Augmentation
• FixMatch: Simplifying Semi-Supervised Learning with Consistency and Confidence
• Affinity and Diversity: Quantifying Mechanisms of Data Augmentation
• Circumventing Outliers of AutoAugment with Knowledge Distillation
• Learning Multiple Layers of Features from Tiny Images

Cited by

• AGVBench: A Reliability-Oriented Benchmark of Data Augmentation for Vein Recognition
• Keep the Core: Adversarial Priors for Significance-Preserving Brain MRI Segmentation
• Experimental insights into data augmentation techniques for deep learning-based multimode fiber imaging: limitations and success
• IPF-RDA: An Information-Preserving Framework for Robust Data Augmentation
• Self Identity Mapping
• Label Smoothing++: Enhanced Label Regularization for Training Neural Networks

Related

• KeepOriginalAugment: Single Image-based Better Information-Preserving Data Augmentation Approach
• Tied-Augment: Controlling Representation Similarity Improves Data Augmentation
• Adversarial Auto-Augment with Label Preservation: A Representation Learning Principle Guided Approach
• Image Data Augmentation Approaches: A Comprehensive Survey and Future directions
• InstructMixup: Instruction-Guided Salient Patch Editing for Robust Data Augmentation