Contrastive learning with semantic consistency constraint

Open Webpage

Huijie Guo, Lei Shi

Published: 01 Jan 2023, Last Modified: 15 May 2025Image Vis. Comput. 2023EveryoneRevisionsBibTeXCC BY-SA 4.0
Abstract: Highlights•Focusing on the semantic inconsistency in contrastive learning.•Semantic consistency regularization alleviates the semantic inconsistency.•Pairwise subspace distance for measuring the distribution difference.•PSD is a plug and play module for contrastive learning methods.