What Can Style Transfer and Paintings Do For Model Robustness?

What Can Style Transfer and Paintings Do For Model Robustness?

Lin, Hubert (Cornell University)
van Zuijlen, M.J.P. (TU Delft Human Information Communication Design)
Pont, S.C. (TU Delft Human Information Communication Design)
Wijntjes, M.W.A. (TU Delft Human Information Communication Design)
Bala, Kavita (Cornell University)

2021

A common strategy for improving model robustness is through data augmentations. Data augmentations encourage models to learn desired invariances, such as invariance to horizontal flipping or small changes in color. Recent work has shown that arbitrary style transfer can be used as a form of data augmentation to encourage invariance to textures by creating painting-like images from photographs. However, a stylized photograph is not quite the same as an artist-created painting. Artists depict perceptually meaningful cues in paintings so that humans can recognize salient components in scenes, an emphasis which is not enforced in style transfer. Therefore, we study how style transfer and paintings differ in their impact on model robustness. First, we investigate the role of paintings as style images for stylization-based data augmentation. We find that style transfer functions well even without paintings as style images. Second, we show that learning from paintings as a form of perceptual data augmentation can improve model robustness. Finally, we investigate the invariances learned from stylization and from paintings, and show that models learn different invariances from these differing forms of data. Our results provide insights into how stylization improves model robustness, and provide evidence that artist-created paintings can be a valuable source of data for model robustness. Code and data are available at: https://github.com/hubertsgithub/style_painting_robustness

http://resolver.tudelft.nl/uuid:e97fd9a6-8a64-4f43-a111-ef3412d30680

https://doi.org/10.1109/CVPR46437.2021.01088

IEEE, Piscataway

978-1-6654-4510-8

Proceedings - 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2021: Proceedings

2021 IEEE/CVF Conference on Computer Vision
and Pattern Recognition, 2021-06-20 → 2021-06-25, Virtual at Nashville, United States

Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 1063-6919

Accepted Author Manuscript

Institutional Repository

conference paper

© 2021 Hubert Lin, M.J.P. van Zuijlen, S.C. Pont, M.W.A. Wijntjes, Kavita Bala