Weakly- and Semi-Supervised Panoptic Segmentation

ECCV 2018 Qizhu LiAnurag ArnabPhilip H. S. Torr

We present a weakly supervised model that jointly performs both semantic- and instance-segmentation -- a particularly relevant problem given the substantial cost of obtaining pixel-perfect annotation for these tasks. In contrast to many popular instance segmentation approaches based on object detectors, our method does not predict any overlapping instances... (read more)

PDF Abstract

Evaluation Results from the Paper


TASK DATASET MODEL METRIC NAME METRIC VALUE GLOBAL RANK COMPARE
Panoptic Segmentation Cityscapes val Dynamically Instantiated Network (ResNet-101) PQ 53.8 # 12
Panoptic Segmentation Cityscapes val Dynamically Instantiated Network (ResNet-101) PQst 62.1 # 9
Panoptic Segmentation Cityscapes val Dynamically Instantiated Network (ResNet-101) PQth 42.5 # 11
Panoptic Segmentation Cityscapes val Dynamically Instantiated Network (ResNet-101) mIoU 79.8 # 1
Panoptic Segmentation Cityscapes val Dynamically Instantiated Network (ResNet-101) AP 28.6 # 10