SOLQ: Segmenting Objects by Learning Queries

Benchmark Model Rank Results
instance-segmentation-on-cocoSOLQ (Swin-L, single scale)#34mask AP: 46.7
instance-segmentation-on-cocoSOLQ (ResNet101, single scale)#61mask AP: 40.9
instance-segmentation-on-cocoSOLQ (ResNet50, single scale)#72mask AP: 39.7
object-detection-on-cocoSOLQ (Swin-L, single scale)#40box mAP: 56.5AP50: 74.6AP75: 60.5APS: 37.6APM: 60APL: 70.6
object-detection-on-cocoSOLQ (ResNet101, single scale)#102box mAP: 48.7
object-detection-on-cocoSOLQ (ResNet50, single scale)#114box mAP: 47.8
object-detection-on-coco-minivalSOLQ (Swin-L, single scale)#211AP50: 74.9AP75: 61.3APL: 71.9