AdaptIS: Adaptive Instance Selection Network

ICCV 2019 Konstantin SofiiukOlga BarinovaAnton Konushin

We present Adaptive Instance Selection network architecture for class-agnostic instance segmentation. Given an input image and a point $(x, y)$, it generates a mask for the object located at $(x, y)$... (read more)

PDF Abstract

Results from the Paper


TASK DATASET MODEL METRIC NAME METRIC VALUE GLOBAL RANK RESULT LEADERBOARD
Panoptic Segmentation Cityscapes val AdaptIS (ResNeXt-101) PQ 62.0 # 4
PQst 64.4 # 5
PQth 58.7 # 3
mIoU 79.2 # 5
AP 36.3 # 8
Panoptic Segmentation Cityscapes val AdaptIS (ResNet-101) PQ 60.6 # 7
PQst 62.9 # 8
PQth 57.5 # 5
mIoU 77.2 # 8
AP 33.9 # 10
Panoptic Segmentation Cityscapes val AdaptIS (ResNet-50) PQ 59.0 # 12
PQst 61.3 # 13
PQth 55.8 # 9
mIoU 75.3 # 11
AP 32.3 # 13
Panoptic Segmentation COCO test-dev AdaptIS (ResNeXt-101) PQ 42.8 # 7
PQst 31.8 # 7
PQth 50.1 # 7
Panoptic Segmentation Mapillary val AdaptIS (ResNeXt-101) PQ 35.9 # 3

Methods used in the Paper