Compactness driven Co-learning for crowd counting and localization

doi:10.1016/j.patcog.2026.113638

Compactness driven Co-learning for crowd counting and localization

2026-03-29

0

PRE

AI

DOI：10.1016/j.patcog.2026.113638

Original

Original request for help

Save

Abstract

En 中文

• A novel approach based on the compactness score to facilitate collaborative learning between confidence and offset predictors for crowd counting and localization. • A novel point-localization algorithm that precisely locates each individual through associated regions of maximum compactness. • Extensive experiments on six crowd benchmarks to validate the effectiveness of our counting and localization approach.

Keywords:

Compactness score

Collaborative learning

Crowd counting

Point-localization

Offset predictors

Compactness driven Co-learning for crowd counting and localization

Abstract

Journal

Organization