Instance Segmentation on LiTS (val)
63.8APMask R-CNN
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| Mask R-CNNBackbone=ResNet-50, Input size=640x640, Training schedule=1x, Multi-scale training=false2022.12 | 63.8 | 83 | 69.9 | 85.7 | 41.9 | |
| SOLOv2Backbone=ResNet-50, Input size=640x640, Training schedule=1x, Multi-scale training=false2022.12 | 63.1 | 80.2 | 66.6 | 89.3 | 36.9 | |
| Box2Mask-TBackbone=ResNet-101, Input size=640x640, Training schedule=2x, Multi-scale training=true2022.12 | 55.3 | 80 | 58.4 | 78 | 32.5 | |
| Box2Mask-TBackbone=ResNet-50, Input size=640x640, Training schedule=2x, Multi-scale training=true2022.12 | 54.9 | 80 | 57.6 | 77.6 | 32.2 | |
| Box2Mask-CBackbone=ResNet-101, Input size=640x640, Training schedule=1x, Multi-scale training=false2022.12 | 52.8 | 79.5 | 56 | 73.8 | 31.7 | |
| Box2Mask-CBackbone=ResNet-50, Input size=640x640, Training schedule=1x, Multi-scale training=false2022.12 | 52.3 | 79.3 | 55.4 | 72.9 | 31.8 | |
| BoxInstBackbone=ResNet-101, Input size=640x640, Training schedule=1x, Multi-scale training=false2022.12 | 50.8 | 79.8 | 52.2 | 73.9 | 27.5 | |
| BoxInstBackbone=ResNet-50, Input size=640x640, Training schedule=1x, Multi-scale training=false2022.12 | 48.9 | 77.3 | 49.8 | 72.9 | 25 |