Are Synthetic Corruptions A Reliable Proxy For Real-World Corruptions?

Shashank Agnihotri, David Schader, Nico Sharei, Mehmet Ege Kaçar, Margret Keuper

Keywords: synthetic corruptions, 2d common corruptions, semantic segmentation.

TL;DR: Synthetic Common Corruptions on real word images serve as a reliable proxy for distriubtion shifts possible in the wild for understanding relative performance.

Abstract: Deep learning (DL) models are widely used in real-world applications but remain vulnerable to distribution shifts, especially due to weather and lighting changes. Collecting diverse real-world data for testing the robustness of DL models is resource-intensive, making synthetic corruptions an attractive alternative for robustness testing. However, are synthetic corruptions a reliable proxy for real-world corruptions? To answer this, we conduct the largest benchmarking study on semantic segmentation models, comparing performance on real-world corruptions and synthetic corruptions datasets. Our results reveal a strong correlation in mean performance, supporting the use of synthetic corruptions for robustness evaluation. We further analyze corruption-specific correlations, providing key insights to understand when synthetic corruptions succeed in representing real-world corruptions. The code and datasets will be released upon acceptance.

Submission Number: 5