Publications
ワークショップ (国際) Mixed Samples Data Augmentation with Replacing Latent Vector Components in Normalizing Flow
Genki Osada, Budrul Ahsan (IBM), Takashi Nishide (University of Tsukuba)
2022 Conference on Neural Information Processing Systems (NeurIPS 2022 First Workshop on Interpolation and Beyond)
2022.11.28
Data augmentation mixing two samples has been acknowledged as an effective regularization method for various deep neural network models. Given that images mixed by popular methods (e.g., MixUp and CutMix) are unnatural to the human eye, we hypothesized that generating more natural images could achieve better performance as data augmentation. To verify this, we propose a new mixing method that synthesizes images in which two source images coexist naturally. Our method performs a mixing operation in latent space through a normalizing flow, and the key is how to mix two latent vectors. We preliminarily observed that there exists a dependency between the dimensions in input space and those in latent space in transformation with normalizing flows. Based on this observation, we designed our mixing scheme in latent space. We show that our method yields visually natural augmented images and improves classification performance.
Paper :
Mixed Samples Data Augmentation with Replacing Latent Vector Components in Normalizing Flow
（外部サイト）