End-to-End Object Detection with Transformers

Benchmark Model Rank Results
panoptic-segmentation-on-coco-minivalDETR-R101 (ResNet-101)#27PQ: 45.1PQst: 37PQth: 50.5RQ: 55.5SQ: 79.9RQst: 46…
panoptic-segmentation-on-coco-minivalPanopticFPN++#29PQ: 44.1PQst: 33.6PQth: 51.0RQ: 53.3SQ: 79.5RQst: 42.1…
object-detection-on-coco-oDETR (ResNet-50)#36Average mAP: 17.1Effective Robustness: -1.82
real-time-object-detection-on-coco-val2017Faster R-CNN-FPN+ (ResNet-50, reported by DETR)#129box AP: 42.0FPS (batch 1, GPU in brackets): 26
object-detection-on-coco-val2017DETR-DC5 (ResNet-101)#316box AP: 44.9AP50: 64.7AP75: 47.7APS: 23.7APM: 49.5APL: 62.3
object-detection-on-coco-val2017Faster RCNN-R101-FPN+#335box AP: 44AP50: 63.9AP75: 47.8APS: 27.2APM: 48.1APL: 56