1 paper
Chang Liu, Giulia Rizzoli, Pietro Zanuttigh +2
Current weakly-supervised incremental learning for semantic segmentation (WILSS) approaches only consider replacing pixel-level annotations with image-level labels, while the train…