TTA-CIFAR-10-C

Mirror of CIFAR-10-C (Hendrycks & Dietterich, ICLR 2019) with a revision pin for reproducible test-time adaptation evaluation.

• Upstream: Zenodo record 2535967
• License: CC BY 4.0 (matches upstream)
• SHA256 of upstream tarball: c72763e101c723b7c507b96205f7e938912a5d587376173b825850cf3cb876a7
• Maintained as part of: TTA-Evaluation-Harness

Citation

@inproceedings{hendrycks2019benchmarking,
  title={Benchmarking Neural Network Robustness to Common Corruptions and Perturbations},
  author={Hendrycks, Dan and Dietterich, Thomas},
  booktitle={ICLR},
  year={2019}
}

Structure

• 15 configs: one per corruption type (gaussian_noise, shot_noise, …, jpeg_compression).
• 5 splits per config: severity_1 through severity_5, 10 000 images each.
• Same 10 000 CIFAR-10 test labels are reused under every corruption/severity.

Usage

from datasets import load_dataset

ds = load_dataset("WNJXYK/TTA-CIFAR-10-C",
                  name="gaussian_noise",
                  split="severity_5",
                  revision="v1.0")

Provenance

This mirror was built by scripts/publish_cifar10c.py in the TTA-Evaluation-Harness repo. No corruption images are re-generated — bytes are copied 1:1 from the upstream .npy files.