Using Videos to Evaluate Image Model Robustness

TL;DR

This paper investigates how different image models handle natural transformations in videos, revealing that more accurate models are generally more robust and that current training techniques do not consistently improve natural robustness.

Contribution

First study of image model robustness to natural video transformations, linking accuracy with robustness and evaluating training methods for brittleness reduction.

Findings

More accurate models are more robust to natural transformations.

Robustness to synthetic color distortions correlates with natural robustness.

Most brittleness in videos is outside adversarial examples.

Abstract

Human visual systems are robust to a wide range of image transformations that are challenging for artificial networks. We present the first study of image model robustness to the minute transformations found across video frames, which we term "natural robustness". Compared to previous studies on adversarial examples and synthetic distortions, natural robustness captures a more diverse set of common image transformations that occur in the natural environment. Our study across a dozen model architectures shows that more accurate models are more robust to natural transformations, and that robustness to synthetic color distortions is a good proxy for natural robustness. In examining brittleness in videos, we find that majority of the brittleness found in videos lies outside the typical definition of adversarial examples (99.9\%). Finally, we investigate training techniques to reduce brittleness and find that no single technique systematically improves natural robustness across twelve tested architectures.