ImageNet-C

ImageNet-C is a benchmark dataset that contains images from the ImageNet dataset corrupted by various types of noise and distortions, used to evaluate the robustness of vision models against these corruptions.

Papers using ImageNet-C (2)

Robustifying Vision Transformer Without Retraining From Scratch By Test-time Class-conditional Feature Alignment2022 · 17 citesUnderstanding The Robustness in Vision Transformers2022 · 34 cites