Optimizing domain-generalizable ReID through non-parametric normalization

Amran Bhuiyan, Aijun An, Jimmy Xiangji Huang, Jialie Shen

Abstract: Highlights•Critical Insights: Non-parametric normalization improves model generalization in ReID.•Novel Framework: DualNormNP uses non-parametric IN and BN to optimize ReID models.•Optimal Layer Integration: Identifying layers for non-parametric IN to optimize gating.•ViT Extension: DualNormNP adapted for ViTs to improve domain-invariant feature extraction.