Fully Convolutional Networks for Panoptic Segmentation

Benchmark Model Rank Results
panoptic-segmentation-on-cityscapes-valPanoptic FCN* (ResNet-FPN)#20PQ: 61.4PQth: 54.8
panoptic-segmentation-on-cityscapes-valPanoptic FCN* (Swin-L, Cityscapes-fine)#26PQst: 70.6PQth: 59.5
panoptic-segmentation-on-cityscapes-valPanoptic FCN* (ResNet-50-FPN)#27PQst: 66.6
panoptic-segmentation-on-coco-minivalPanoptic FCN* (ResNet-50-FPN)#26PQ: 44.3PQst: 35.6PQth: 50RQ: 53SQ: 80.7RQst: 43.5
panoptic-segmentation-on-coco-minivalPanoptic FCN* (Swin-L, single-scale)#31PQth: 58.5RQ: 61.6SQ: 83.2RQst: 51.1RQth: 68.6SQst: 81.1
panoptic-segmentation-on-coco-test-devPanoptic FCN* (Swin-L)#10PQ: 52.7PQth: 59.4
panoptic-segmentation-on-coco-test-devPanoptic FCN*++ (DCN-101-FPN)#18PQ: 47.5PQst: 38.2PQth: 53.7
panoptic-segmentation-on-mapillary-valPanoptic FCN* (Swin-L, single-scale)#3PQ: 45.7PQst: 52.1PQth: 40.8
panoptic-segmentation-on-mapillary-valPanoptic FCN* (ResNet-FPN)#8PQ: 36.9PQth: 32.9
panoptic-segmentation-on-mapillary-valPanoptic FCN* (ResNet-50-FPN)#10PQst: 42.3