Learning Instance Occlusion for Panoptic Segmentation
This addresses a specific bottleneck in panoptic segmentation for computer vision applications, offering an incremental improvement over existing fusion methods.
The paper tackles the problem of incorrect occlusion relationships in panoptic segmentation by proposing OCFusion, a lightweight branch that models instance mask overlaps as binary relations, achieving state-of-the-art results on COCO and competitive performance on Cityscapes.
Panoptic segmentation requires segments of both "things" (countable object instances) and "stuff" (uncountable and amorphous regions) within a single output. A common approach involves the fusion of instance segmentation (for "things") and semantic segmentation (for "stuff") into a non-overlapping placement of segments, and resolves overlaps. However, instance ordering with detection confidence do not correlate well with natural occlusion relationship. To resolve this issue, we propose a branch that is tasked with modeling how two instance masks should overlap one another as a binary relation. Our method, named OCFusion, is lightweight but particularly effective in the instance fusion process. OCFusion is trained with the ground truth relation derived automatically from the existing dataset annotations. We obtain state-of-the-art results on COCO and show competitive results on the Cityscapes panoptic segmentation benchmark.