Object Localization

Acquiring count annotations generally requires less human effort than point-level and bounding box annotations. Thus, we propose the …