MetaShift
Introduced by Liang et al. in MetaShift: A Dataset of Datasets for Evaluating Contextual Distribution Shifts and Training Conflicts
MetaShift is a collection of 12,868 sets of natural images across 410 classes. It can be used to benchmark and evaluate how robust machine learning models are to data shifts.
Papers
| Paper | Code | Results | Date | Stars |
|---|
