Region Mixup

TL;DR

This paper proposes a region-based extension of mixup data augmentation that combines parts of images to improve generalization in visual recognition tasks.

Contribution

It introduces a novel region mixup method that enhances data augmentation by blending image regions instead of whole images.

Findings

Improved accuracy in visual recognition benchmarks

Enhanced model robustness to variations

Simple yet effective augmentation technique

Abstract

This paper introduces a simple extension of mixup (Zhang et al., 2018) data augmentation to enhance generalization in visual recognition tasks. Unlike the vanilla mixup method, which blends entire images, our approach focuses on combining regions from multiple images.