Scale Adaptive Feature Pyramid Networks for 2D Object Detection

Table 2

Accuracies in mAP [%] of the faster R-CNN [5] without the FPN, with the (original) FPN, and the proposed SAFPN.

	W/O FPN	FPN	SAFPN
			L0.5	L1	L2	(L2)^2	L∞
Faster R-CNN	74.6	76.1	76.1	78.1	78.3	78.3	75.3
RetinaNet		73.3	72.0	74.2	74.1	74.2	73.1

Accuracies for RetinaNet [10] are also shown with the FPN and SAFPN. Training is done using the Pascal VOC 2007 train dataset.