CoLo-CAM: Class activation mapping for object co-localization in weakly-labeled unconstrained videos

Soufiane Belharbi, Shakeeb Murtaza, Marco Pedersoli, Ismail Ben Ayed, Luke McCaffrey, Eric Granger

Abstract: Highlights•CoLo-CAM method for weakly-supervised object localization in unconstrained videos.•Color cues are employed during training to co-locate objects over frames.•CoLo-CAM builds a fully connected graph over pixels from all frames.•It achieves new state-of-the-art localization performance on YouTube-Objects data.•Results, ablations, code, and demonstrative videos are provided for validation.