Instance Segmentation on COCO non-VOC categories
39.7mAPContrastMask
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| ContrastMaskBackbone=ResNeXt-101, Sche.=3x, Layers.=42022.03 | 39.7 | 66.2 | 42.3 | 19.2 | 39.3 | 53.6 | |
| ContrastMaskBackbone=ResNet-101, Sche.=3x, Layers.=42022.03 | 38.4 | 64.5 | 39.8 | 18.4 | 38.1 | 52.6 | |
| ContrastMaskBackbone=ResNeXt-50, Sche.=3x, Layers.=42022.03 | 37.6 | 63.8 | 38.9 | 18.1 | 36.6 | 51.3 | |
| OPMaskBackbone=ResNet-101, Sche.=130k, Layers.=72022.03 | 37.1 | 62.5 | 38.4 | 16.9 | 36 | 50.5 | |
| ContrastMaskBackbone=ResNet-50, Sche.=3x, Layers.=42022.03 | 37 | 63 | 38.6 | 18.3 | 36.4 | 50.2 | |
| CPMaskBackbone=ResNet-101, Sche.=3x, Layers.=42022.03 | 36.8 | 60.5 | 38.6 | 17.6 | 37.1 | 51.5 | |
| ContrastMaskBackbone=ResNet-101, Sche.=1x, Layers.=42022.03 | 36.6 | 62.2 | 37.7 | 17.5 | 36.5 | 50.1 | |
| OPMaskBackbone=ResNet-50, Sche.=130k, Layers.=72022.03 | 36.5 | 62.5 | 37.4 | 17.3 | 34.8 | 49.8 | |
| ShapeMask (NAS-FPN)Backbone=ResNet-101, Sche.=3x, Layers.=82022.03 | 35.7 | 60.3 | 36.6 | 18.3 | 40.5 | 47.3 | |
| ShapePropBackbone=ResNet-101, Sche.=1x, Layers.=42022.03 | 35.5 | 60.5 | 36.7 | 15.6 | 33.8 | 50.3 | |
| ContrastMaskBackbone=ResNet-50, Sche.=1x, Layers.=42022.03 | 35.1 | 60.8 | 35.7 | 17.2 | 34.7 | 47.7 | |
| ContrastMaskBackbone=ResNeXt-101, Sche.=3x, Layers.=42022.03 | 35 | 56.4 | 36.9 | 18.6 | 38.9 | 50.5 | |
| Mask R-CNNSupervision=Full supervision2021.04 | 34.4 | 55.2 | 36.3 | 18.6 | 39.1 | 47.9 | |
| ShapePropBackbone=ResNet-50, Sche.=1x, Layers.=42022.03 | 34.4 | 59.6 | 35.2 | 13.5 | 32.9 | 48.6 | |
| ContrastMaskBackbone=ResNet-101, Sche.=3x, Layers.=42022.03 | 34.3 | 54.7 | 36.6 | 17.5 | 38.4 | 50 | |
| CPMaskBackbone=ResNet-101, Sche.=3x, Layers.=42022.03 | 34 | 53.7 | 36.5 | 18.5 | 38.9 | 47.4 | |
| ContrastMaskBackbone=ResNeXt-50, Sche.=3x, Layers.=42022.03 | 33.4 | 54.2 | 35.8 | 17.7 | 37.4 | 48.5 | |
| ShapeMaskBackbone=ResNet-101, Sche.=1x, Layers.=82022.03 | 33.3 | 56.9 | 34.3 | 17.1 | 38.1 | 45.4 | |
| ShapeMask (NAS-FPN)Backbone=ResNet-101, Sche.=3x, Layers.=82022.03 | 33.2 | 53.1 | 35 | 15.5 | 40.2 | 43.3 | |
| OPMaskBackbone=ResNet-101, Sche.=130k, Layers.=72022.03 | 33.2 | 53.5 | 35.2 | 17.2 | 37.1 | 46.9 | |
| ContrastMaskBackbone=ResNet-50, Sche.=3x, Layers.=42022.03 | 32.9 | 52.5 | 35.4 | 16.6 | 37.1 | 47.3 | |
| OursSupervision=VOC Masks2021.04 | 32.7 | 52.5 | 33.1 | 15.9 | 38.5 | 41.3 | |
| ContrastMaskBackbone=ResNet-101, Sche.=1x, Layers.=42022.03 | 32.4 | 52.1 | 34.8 | 15.2 | 36.7 | 47.3 | |
| OPMaskBackbone=ResNet-50, Sche.=130k, Layers.=72022.03 | 31.9 | 52.2 | 33.7 | 16.3 | 35.2 | 46.5 | |
| ShapePropBackbone=ResNet-101, Sche.=1x, Layers.=42022.03 | 31.9 | 52.1 | 33.7 | 14.2 | 35.9 | 46.5 | |
| ContrastMaskBackbone=ResNet-50, Sche.=1x, Layers.=42022.03 | 30.9 | 50.3 | 32.9 | 15.2 | 34.6 | 44.3 | |
| ShapePropBackbone=ResNet-50, Sche.=1x, Layers.=42022.03 | 30.4 | 51.2 | 31.8 | 14.3 | 34.2 | 44.7 | |
| ShapeMaskSupervision=VOC Masks + non-VOC Boxes2021.04 | 30.2 | 49.3 | 31.5 | 16.1 | 38.2 | 38.4 | |
| ShapeMaskBackbone=ResNet-101, Sche.=1x, Layers.=82022.03 | 30.2 | 49.3 | 31.5 | 16.1 | 38.2 | 28.4 | |
| Mask* R-CNNBackbone=ResNet-101, Sche.=1x, Layers.=42022.03 | 29.5 | 52.4 | 29.2 | 13.4 | 30.2 | 41 | |
| Mask* R-CNNBackbone=ResNet-50, Sche.=1x, Layers.=42022.03 | 28.9 | 52.2 | 28.6 | 12.1 | 29 | 40.6 | |
| CPMaskBackbone=ResNet-50, Sche.=1x, Layers.=42022.03 | 28.8 | 46.1 | 30.6 | 12.4 | 33.1 | 43.4 | |
| Mask R-CNN (Baseline)Backbone=ResNet-50, Sche.=1x, Layers.=42022.03 | 23.9 | 42.9 | 23.5 | 11.6 | 24.3 | 33.7 | |
| Mask^X R-CNNSupervision=VOC Masks + non-VOC Boxes2021.04 | 23.8 | 42.9 | 23.5 | 12.7 | 28.1 | 33.5 | |
| Mask* R-CNNBackbone=ResNet-101, Sche.=1x, Layers.=42022.03 | 23.8 | 42.9 | 23.5 | 12.7 | 28.1 | 33.5 | |
| Mask* R-CNNBackbone=ResNet-50, Sche.=1x, Layers.=42022.03 | 23.7 | 43.1 | 23.5 | 12.4 | 27.6 | 32.9 | |
| Mask GrabCutBackbone=ResNet-101, Sche.=1x2022.03 | 19.7 | 39.7 | 17 | 6.4 | 21.2 | 35.8 | |
| Mask GrabCutBackbone=ResNet-101, Sche.=1x2022.03 | 19.6 | 46.1 | 14.3 | 5.1 | 16 | 32.4 | |
| Mask GrabCutBackbone=ResNet-50, Sche.=1x2022.03 | 19.5 | 46.2 | 14.2 | 4.7 | 15.9 | 32 | |
| Mask GrabCutBackbone=ResNet-50, Sche.=1x2022.03 | 19.5 | 39.2 | 17 | 6.5 | 20.9 | 34.3 | |
| Mask R-CNN (Baseline)Backbone=ResNet-50, Sche.=1x, Layers.=42022.03 | 19.2 | 36.4 | 18.4 | 11.5 | 23.3 | 24.4 |