Subdivided Mask Dispersion Framework for semi-supervised semantic segmentation

Yooseung Wang, Jaehyuk Jang, Changick Kim

Abstract: Highlights•We proposed a novel Subdivided Mask Dispersion Framework (SMDF) for semi-supervised semantic segmentation.•We proposed a Subdivided White Mask Disperser (SWMD) that generates small and uniformly distributed bounding boxes.•SMDF outperforms all the semi-supervised semantic segmentation approaches on PASCAL and Extended SBD datasets.•We substantiated the effectiveness of SMDF in different backbone networks and resolutions.