VITA: A Multi-Source Vicinal Transfer Augmentation Method for Out-of-Distribution Generalization

VITA generates diverse on-manifold samples through multi-source vicinal transfer to improve robustness to image corruptions.

Corruption Invariant Person Re-Identification

A talk (in Chinese) about corruption invariant person re-identification and understanding model robustness against image corruptions.

Benchmarks for Corruption Invariant Person Re-Identification

The first comprehensive study of five ReID corruption benchmarks compares transformer and CNN robustness and examines cross-dataset generalization.