Weakly supervised object localization via foreground generation with foreground–background constraints
Abstract: Highlights•Introducing a novel foreground prediction map generator for weakly supervised object localization.•Introducing a self-guided strategy to guide foreground and background in FPM.•Introducing an adversarial learning strategy to balance foregrond and background in FPM.
External IDs:doi:10.1016/j.eswa.2024.126273
Loading