Mask R-CNN

Reported on 61 benchmarks across 10 tasks · 2 papers · 23 SOTA

Note: results are matched by exact model name. Different papers may use the same name for different model variants.

Computer Vision30 results

Object LocalizationonGRIT
Localization (ablation)· 2017-03-20
44.7
best: 67 (Unified-IOXL)
SOTA
Mask R-CNN arXiv:1703.06870
Object LocalizationonGRIT
Localization (test)· 2017-03-20
45.1
best: 67.1 (Unified-IOXL)
SOTA
Mask R-CNN arXiv:1703.06870
Pose EstimationonCOCO
Test AP· 2017-03-20
63.1
SOTA
Mask R-CNN arXiv:1703.06870
Pose EstimationonCOCO
Validation AP· 2017-03-20
69.2
SOTA
Mask R-CNN arXiv:1703.06870
Pose EstimationonOCHuman
AP50· 2017-03-20
33.2
best: 89.7 (MIPNet (gt-bb))
SOTA
Mask R-CNN arXiv:1703.06870
Pose EstimationonOCHuman
AP75· 2017-03-20
24.5
best: 80.1 (MIPNet (gt-bb))
SOTA
Mask R-CNN arXiv:1703.06870
Instance SegmentationonBDD100K val
AP· 2017-03-20
20.5
best: 23.6 (Mask Transfiner)
SOTA
Mask R-CNN arXiv:1703.06870
Human ParsingonMHP v2.0
AP 0.5· 2017-03-20
14.9
best: 51.2 (UniParser)
SOTA
Mask R-CNN arXiv:1703.06870
Object SegmentationonGRIT
Segmentation (ablation)· 2017-03-20
26.2
best: 56.3 (Unified-IOXL)
SOTA
Mask R-CNN arXiv:1703.06870
Object SegmentationonGRIT
Segmentation (test)· 2017-03-20
26.2
best: 56.5 (Unified-IOXL)
SOTA
Mask R-CNN arXiv:1703.06870
Multi-Person Pose EstimationonOCHuman
AP50· 2017-03-20
33.2
best: 89.7 (MIPNet (gt-bb))
SOTA
Mask R-CNN arXiv:1703.06870
Multi-Person Pose EstimationonOCHuman
AP75· 2017-03-20
24.5
best: 80.1 (MIPNet (gt-bb))
SOTA
Mask R-CNN arXiv:1703.06870
Multi-Person Pose EstimationonOCHuman
Validation AP· 2017-03-20
20.2
best: 74.1 (MIPNet (gt-bb))
SOTA
Mask R-CNN arXiv:1703.06870
Pose EstimationonStanfordExtra
PCK@0.1· 2021-07-31
50.77
best: 78.65 (8 Stacked Hourglass Network)
SyDog: A Synthetic Dog Dataset for Improved 2D Pose Estimation arXiv:2108.00249
Animal Pose EstimationonStanfordExtra
PCK@0.1· 2021-07-31
50.77
best: 78.65 (8 Stacked Hourglass Network)
SyDog: A Synthetic Dog Dataset for Improved 2D Pose Estimation arXiv:2108.00249
Pose EstimationonCOCO test-dev
AP50· 2017-03-20
87.3
best: 95 (ViTPose (ViTAE-G, ensemble))
Mask R-CNN arXiv:1703.06870
Pose EstimationonCOCO test-dev
AP75· 2017-03-20
68.7
best: 88.2 (ViTPose (ViTAE-G, ensemble))
Mask R-CNN arXiv:1703.06870
Pose EstimationonCOCO test-dev
APL· 2017-03-20
71.4
best: 86.5 (PoseBH-H)
Mask R-CNN arXiv:1703.06870
Pose EstimationonCOCO test-dev
APM· 2017-03-20
57.8
best: 83.8 (4xRSN-50 (ensemble))
Mask R-CNN arXiv:1703.06870
Pose EstimationonCrowdPose
AP Easy· 2017-03-20
69.4
best: 88.8 (RTMO-l)
Mask R-CNN arXiv:1703.06870
Pose EstimationonCrowdPose
AP Hard· 2017-03-20
45.8
best: 466 (DETRPose-N)
Mask R-CNN arXiv:1703.06870
Pose EstimationonCrowdPose
AP Medium· 2017-03-20
57.9
best: 566 (DETRPose-N)
Mask R-CNN arXiv:1703.06870
Pose EstimationonCrowdPose
mAP @0.5:0.95· 2017-03-20
57.2
best: 83.8 (RTMO-l)
Mask R-CNN arXiv:1703.06870
Pose EstimationonOCHuman
Validation AP· 2017-03-20
20.2
best: 92.8 (ViTPose (ViTAE-G, GT bounding boxes))
Mask R-CNN arXiv:1703.06870
Multi-Person Pose EstimationonCrowdPose
AP Easy· 2017-03-20
69.4
best: 88.8 (RTMO-l)
Mask R-CNN arXiv:1703.06870
Multi-Person Pose EstimationonCrowdPose
AP Hard· 2017-03-20
45.8
best: 466 (DETRPose-N)
Mask R-CNN arXiv:1703.06870
Multi-Person Pose EstimationonCrowdPose
AP Medium· 2017-03-20
57.9
best: 566 (DETRPose-N)
Mask R-CNN arXiv:1703.06870
Multi-Person Pose EstimationonCrowdPose
mAP @0.5:0.95· 2017-03-20
57.2
best: 83.8 (RTMO-l)
Mask R-CNN arXiv:1703.06870
Instance SegmentationonMSCOCO
Frame (fps)· uses extra data
5
best: 172.7 (YolactEdge)
Instance SegmentationonMSCOCO
mask AP· uses extra data
34.6
best: 44.6 (RTMDet-Ins-x)

Methodology14 results

3DonCOCO
Test AP· 2017-03-20
63.1
SOTA
Mask R-CNN arXiv:1703.06870
3DonCOCO
Validation AP· 2017-03-20
69.2
SOTA
Mask R-CNN arXiv:1703.06870
3DonOCHuman
AP50· 2017-03-20
33.2
best: 89.7 (MIPNet (gt-bb))
SOTA
Mask R-CNN arXiv:1703.06870
3DonOCHuman
AP75· 2017-03-20
24.5
best: 80.1 (MIPNet (gt-bb))
SOTA
Mask R-CNN arXiv:1703.06870
3DonStanfordExtra
PCK@0.1· 2021-07-31
50.77
best: 78.65 (8 Stacked Hourglass Network)
SyDog: A Synthetic Dog Dataset for Improved 2D Pose Estimation arXiv:2108.00249
3DonCOCO test-dev
AP50· 2017-03-20
87.3
best: 95 (ViTPose (ViTAE-G, ensemble))
Mask R-CNN arXiv:1703.06870
3DonCOCO test-dev
AP75· 2017-03-20
68.7
best: 88.2 (ViTPose (ViTAE-G, ensemble))
Mask R-CNN arXiv:1703.06870
3DonCOCO test-dev
APL· 2017-03-20
71.4
best: 86.5 (PoseBH-H)
Mask R-CNN arXiv:1703.06870
3DonCOCO test-dev
APM· 2017-03-20
57.8
best: 83.8 (4xRSN-50 (ensemble))
Mask R-CNN arXiv:1703.06870
3DonCrowdPose
AP Easy· 2017-03-20
69.4
best: 88.8 (RTMO-l)
Mask R-CNN arXiv:1703.06870
3DonCrowdPose
AP Hard· 2017-03-20
45.8
best: 466 (DETRPose-N)
Mask R-CNN arXiv:1703.06870
3DonCrowdPose
AP Medium· 2017-03-20
57.9
best: 566 (DETRPose-N)
Mask R-CNN arXiv:1703.06870
3DonCrowdPose
mAP @0.5:0.95· 2017-03-20
57.2
best: 83.8 (RTMO-l)
Mask R-CNN arXiv:1703.06870
3DonOCHuman
Validation AP· 2017-03-20
20.2
best: 92.8 (ViTPose (ViTAE-G, GT bounding boxes))
Mask R-CNN arXiv:1703.06870

Audio14 results

1 Image, 2*2 StitchionCOCO
Test AP· 2017-03-20
63.1
SOTA
Mask R-CNN arXiv:1703.06870
1 Image, 2*2 StitchionCOCO
Validation AP· 2017-03-20
69.2
SOTA
Mask R-CNN arXiv:1703.06870
1 Image, 2*2 StitchionOCHuman
AP50· 2017-03-20
33.2
best: 89.7 (MIPNet (gt-bb))
SOTA
Mask R-CNN arXiv:1703.06870
1 Image, 2*2 StitchionOCHuman
AP75· 2017-03-20
24.5
best: 80.1 (MIPNet (gt-bb))
SOTA
Mask R-CNN arXiv:1703.06870
1 Image, 2*2 StitchionStanfordExtra
PCK@0.1· 2021-07-31
50.77
best: 78.65 (8 Stacked Hourglass Network)
SyDog: A Synthetic Dog Dataset for Improved 2D Pose Estimation arXiv:2108.00249
1 Image, 2*2 StitchionCOCO test-dev
AP50· 2017-03-20
87.3
best: 95 (ViTPose (ViTAE-G, ensemble))
Mask R-CNN arXiv:1703.06870
1 Image, 2*2 StitchionCOCO test-dev
AP75· 2017-03-20
68.7
best: 88.2 (ViTPose (ViTAE-G, ensemble))
Mask R-CNN arXiv:1703.06870
1 Image, 2*2 StitchionCOCO test-dev
APL· 2017-03-20
71.4
best: 86.5 (PoseBH-H)
Mask R-CNN arXiv:1703.06870
1 Image, 2*2 StitchionCOCO test-dev
APM· 2017-03-20
57.8
best: 83.8 (4xRSN-50 (ensemble))
Mask R-CNN arXiv:1703.06870
1 Image, 2*2 StitchionCrowdPose
AP Easy· 2017-03-20
69.4
best: 88.8 (RTMO-l)
Mask R-CNN arXiv:1703.06870
1 Image, 2*2 StitchionCrowdPose
AP Hard· 2017-03-20
45.8
best: 466 (DETRPose-N)
Mask R-CNN arXiv:1703.06870
1 Image, 2*2 StitchionCrowdPose
AP Medium· 2017-03-20
57.9
best: 566 (DETRPose-N)
Mask R-CNN arXiv:1703.06870
1 Image, 2*2 StitchionCrowdPose
mAP @0.5:0.95· 2017-03-20
57.2
best: 83.8 (RTMO-l)
Mask R-CNN arXiv:1703.06870
1 Image, 2*2 StitchionOCHuman
Validation AP· 2017-03-20
20.2
best: 92.8 (ViTPose (ViTAE-G, GT bounding boxes))
Mask R-CNN arXiv:1703.06870

Medical3 results

Medical Image SegmentationonCell17
Dice· 2017-03-20
0.707
best: 0.7088 (Cell R-CNN)
SOTA
Mask R-CNN arXiv:1703.06870
Medical Image SegmentationonCell17
F1-score· 2017-03-20
0.8004
best: 0.8216 (Cell R-CNN)
SOTA
Mask R-CNN arXiv:1703.06870
Medical Image SegmentationonCell17
Hausdorff· 2017-03-20
12.6723
best: 25.9102 (FnsNet)
Mask R-CNN arXiv:1703.06870