Recursively Refined R-CNN: Instance Segmentation with Self-RoI Rebalancing

Open paper
Benchmark Model Rank Results
instance-segmentation-on-coco-minivalR3-CNN (ResNet-50-FPN, DCN)#69mask AP: 40.4AP50: 61.3AP75: 44APL: 56.1APM: 43.6APS: 22.3
instance-segmentation-on-coco-minivalR3-CNN (ResNet-50-FPN, GC-Net)#73mask AP: 40.2AP50: 61.1AP75: 43.5APM: 42.8APS: 22.6
instance-segmentation-on-coco-minivalR3-CNN (ResNet-50-FPN, GRoIE)#75mask AP: 39.1AP50: 58.8AP75: 42.3APL: 54.3APM: 42.1APS: 20.7
instance-segmentation-on-coco-minivalR3-CNN (ResNet-50-FPN)#79mask AP: 38.2AP50: 58AP75: 41.4APL: 52.8APM: 41APS: 20.4
object-detection-on-coco-minivalR3-CNN (ResNet-50-FPN, DCN)#123box AP: 44.8AP50: 64.3AP75: 48.9APS: 26.6APM: 48.3APL: 59.6
object-detection-on-coco-minivalR3-CNN (ResNet-50-FPN, GC-Net)#133box AP: 44.3AP50: 64.1AP75: 48.4APS: 27APM: 47.1APL: 58.9
object-detection-on-coco-minivalR3-CNN (ResNet-50-FPN)#158box AP: 42AP50: 61AP75: 46.3APS: 24.5APM: 45.2APL: 55.7
object-detection-on-coco-minivalR3-CNN (ResNet-50-FPN, GRoIE)#216AP50: 61.2AP75: 45.6APS: 24.4

Rank counts only results with a code link.