CBNet: A Composite Backbone Network Architecture for Object Detection

Open paper
Benchmark Model Rank Results
instance-segmentation-on-cocoCBNetV2 (EVA02, single-scale)#2mask AP: 56.1AP50: 80.3AP75: 62.1APS: 39.7APM: 59.3APL: 70.9
instance-segmentation-on-cocoCBNetV2 (Dual-Swin-L HTC, multi-scale)#15mask AP: 52.3
instance-segmentation-on-cocoCBNetV2 (Dual-Swin-L HTC, single-scale)#17mask AP: 51.6
instance-segmentation-on-coco-minivalCBNetV2 (Dual-Swin-L HTC, multi-scale)#17mask AP: 51.8
instance-segmentation-on-coco-minivalCBNetV2 (Dual-Swin-L HTC, multi-scale)#18mask AP: 51
object-detection-on-cocoCBNetV2 (Dual-Swin-L HTC, multi-scale)#29box mAP: 60.1
object-detection-on-cocoCBNetV2 (Dual-Swin-L HTC, single-scale)#31box mAP: 59.4
object-detection-on-coco-minivalCBNetV2 (Dual-Swin-L HTC, multi-scale)#29box AP: 59.6
object-detection-on-coco-minivalCBNetV2 (Dual-Swin-L HTC, multi-scale)#32box AP: 59.1
object-detection-on-coco-oCBNetV2 (Swin-L)#6Average mAP: 39.0Effective Robustness: 12.36

Rank counts only results with a code link.