Table 1.

The object-detection performance comparison on COCO val2017. The best performance is highlighted in bold format.

Model	$\mathit{AP}$	${\mathit{AP}}_{50}$	${\mathit{AP}}_{75}$	${\mathit{AP}}_{\mathit{S}}$	${\mathit{AP}}_{\mathit{M}}$	${\mathit{AP}}_{\mathit{L}}$
Faster-RCNN [3]	42.0	62.1	45.5	26.6	45.4	53.4
RetinaNet [10]	40.8	61.1	44.1	24.1	44.2	51.2
FCOS [27]	41.0	59.8	44.1	26.2	44.6	52.2
TSP-FCOS [28]	43.1	62.3	47.0	26.6	46.8	55.9
TSP-RCNN [28]	43.8	63.3	48.3	28.6	46.9	55.7
DETR [11]	42.0	62.4	44.2	20.5	45.8	61.1
UP-DETR [29]	42.8	63.0	45.3	20.8	47.1	61.7
Anchor-DETR [15]	42.1	63.1	44.9	22.3	46.2	60.0
Conditional-DETR [14]	43.0	64.0	45.7	22.7	46.7	61.5
Deformable-DETR [12]	43.8	62.6	47.7	26.4	47.1	58.0
Ours	44.7	63.9	48.5	27.0	47.5	59.1