Mask R-CNN

ICCV 2017 Kaiming HeGeorgia GkioxariPiotr DollárRoss Girshick

We present a conceptually simple, flexible, and general framework for object instance segmentation. Our approach efficiently detects objects in an image while simultaneously generating a high-quality segmentation mask for each instance... (read more)

PDF Abstract

Code


Evaluation Results from the Paper


 SOTA for Instance Segmentation on Cityscapes test (using extra training data)

     Get a GitHub badge
TASK DATASET MODEL METRIC NAME METRIC VALUE GLOBAL RANK USES EXTRA
TRAINING DATA
COMPARE
Nuclear Segmentation Cell17 Mask R-CNN F1-score 0.8004 # 3
Nuclear Segmentation Cell17 Mask R-CNN Dice 0.7070 # 3
Nuclear Segmentation Cell17 Mask R-CNN Hausdorff 12.6723 # 3
Instance Segmentation Cityscapes test Mask R-CNN + Coco Average Precision 32.0 # 1
Instance Segmentation Cityscapes test Mask R-CNN Average Precision 26.2 # 2
Keypoint Detection COCO Mask R-CNN Validation AP 69.2 # 6
Keypoint Detection COCO Mask R-CNN Test AP 63.1 # 9
Instance Segmentation COCO minival Mask R-CNN (ResNet-101-FPN) mask AP 35.4 # 11
Object Detection COCO minival Mask R-CNN (ResNeXt-101-FPN) box AP 36.7 # 44
Object Detection COCO minival Mask R-CNN (ResNeXt-101-FPN) AP50 59.5 # 20
Object Detection COCO minival Mask R-CNN (ResNeXt-101-FPN) AP75 38.9 # 30
Object Detection COCO minival Mask R-CNN (ResNet-101-FPN) box AP 40.0 # 32
Object Detection COCO minival Mask R-CNN (ResNet-50-FPN) box AP 37.7 # 43
Instance Segmentation COCO minival Mask R-CNN (ResNet-50-FPN) mask AP 33.6 # 12
Instance Segmentation COCO minival Mask R-CNN (ResNeXt-101-FPN) mask AP 36.7 # 8
Instance Segmentation COCO test-dev Mask R-CNN (ResNeXt-101-FPN) mask AP 37.1% # 10
Instance Segmentation COCO test-dev Mask R-CNN (ResNeXt-101-FPN) AP50 60 # 1
Instance Segmentation COCO test-dev Mask R-CNN (ResNeXt-101-FPN) AP75 39.4 # 1
Instance Segmentation COCO test-dev Mask R-CNN (ResNeXt-101-FPN) APS 16.9 # 1
Instance Segmentation COCO test-dev Mask R-CNN (ResNeXt-101-FPN) APM 39.9 # 1
Instance Segmentation COCO test-dev Mask R-CNN (ResNeXt-101-FPN) APL 53.5 # 1
Object Detection COCO test-dev Mask R-CNN (ResNeXt-101-FPN) box AP 39.8 # 38
Object Detection COCO test-dev Mask R-CNN (ResNeXt-101-FPN) AP50 62.3 # 28
Object Detection COCO test-dev Mask R-CNN (ResNeXt-101-FPN) AP75 43.4 # 36
Object Detection COCO test-dev Mask R-CNN (ResNeXt-101-FPN) APS 22.1 # 39
Object Detection COCO test-dev Mask R-CNN (ResNeXt-101-FPN) APM 43.2 # 29
Object Detection COCO test-dev Mask R-CNN (ResNeXt-101-FPN) APL 51.2 # 38
Object Detection COCO test-dev Mask R-CNN (ResNet-101-FPN) box AP 38.2 # 45
Object Detection COCO test-dev Mask R-CNN (ResNet-101-FPN) AP50 60.3 # 36
Object Detection COCO test-dev Mask R-CNN (ResNet-101-FPN) AP75 41.7 # 44
Object Detection COCO test-dev Mask R-CNN (ResNet-101-FPN) APS 20.1 # 50
Object Detection COCO test-dev Mask R-CNN (ResNet-101-FPN) APM 41.1 # 38
Object Detection COCO test-dev Mask R-CNN (ResNet-101-FPN) APL 50.2 # 41
Instance Segmentation COCO test-dev Mask R-CNN (ResNet-101-FPN) mask AP 35.7% # 11
Multi-Human Parsing MHP v1.0 Mask R-CNN AP 0.5 52.68% # 2
Multi-Human Parsing MHP v2.0 Mask R-CNN AP 0.5 14.90% # 3