APR-SP + DeepAugment (ResNet-50)

Reported on 2 benchmarks across 2 tasks · 1 paper

Note: results are matched by exact model name. Different papers may use the same name for different model variants.

Methodology1 result

Domain AdaptationonImageNet-C
mean Corruption Error (mCE)· 2021-08-19
57.5
best: 22 (EfficientNet-L2+RPL)
Amplitude-Phase Recombination: Rethinking Robustness of Convolutional Neural Networks in Frequency Domain arXiv:2108.08487

Computer Vision1 result

Domain GeneralizationonImageNet-C
mean Corruption Error (mCE)· 2021-08-19
57.5
best: 28.2 (DINOv2 (ViT-g/14, frozen model, linear eval))
Amplitude-Phase Recombination: Rethinking Robustness of Convolutional Neural Networks in Frequency Domain arXiv:2108.08487