Colorful Cutout: Enhancing Image Data Augmentation with Curriculum Learning

TL;DR

The paper addresses the gap in considering data difficulty during image augmentation by introducing colorful cutout, a curriculum-based augmentation that colorizes the erased region and progressively increases noise through sub-region division. The approach defines a curriculum via the number of sub-regions as $2^{N_{\text{epoch}}}$, with $N_{\text{epoch}}=0$ initially, and demonstrates its efficacy across CIFAR-10, CIFAR-100, and Tiny ImageNet on diverse architectures including ResNet50, EfficientNet-B0, and ViT-B/16. Results show that colorful cutout outperforms traditional augmentation methods such as Cutout, Mixup, and CutMix, with ablation indicating the curriculum component is key to the gains. This work suggests that curriculum data augmentation is a promising direction for improving generalization in computer vision and invites future exploration of applying curriculum concepts to other augmentations and soft-label strategies.

Abstract

Data augmentation is one of the regularization strategies for the training of deep learning models, which enhances generalizability and prevents overfitting, leading to performance improvement. Although researchers have proposed various data augmentation techniques, they often lack consideration for the difficulty of augmented data. Recently, another line of research suggests incorporating the concept of curriculum learning with data augmentation in the field of natural language processing. In this study, we adopt curriculum data augmentation for image data augmentation and propose colorful cutout, which gradually increases the noise and difficulty introduced in the augmented image. Our experimental results highlight the possibility of curriculum data augmentation for image data. We publicly released our source code to improve the reproducibility of our study.

Figures (2)

• Figure 1: As the training procedure progresses, colorful cutout introduces more complex and difficult noise into augmented images.
• Figure 2: An example of our proposed colorful cutout compared to previous data augmentation methods.