Journal of Remote Sensing / 2021 / Article / Tab 3

Research Article

Feature Enhancement Network for Object Detection in Optical Remote Sensing Images

Table 3

Comparison of FENet and the state-of-the-art methods on the DOTA test set. FR and MR indicate the Faster R-CNN [14] and Mask R-CNN [59] methods, respectively.

MethodPLBDBRGTFSVLVSHTCBCSTSBFRAHASPHCmAP

[19]88.7075.1052.6059.2069.4078.8084.5090.6081.3082.6052.5062.1076.6066.3060.1072.00
EFR [62]88.3683.9045.7867.2476.8077.1585.3590.7785.5575.7754.6460.7671.4077.9060.9473.49
IoU-adaptive R-CNN [63]88.6280.2253.1866.9776.3072.5984.0790.6680.9576.2457.1266.6584.0866.3656.8572.72
FMSSD [64]89.1181.5148.2267.9469.2373.5676.8790.7182.6773.3352.6567.5272.3780.5760.1572.43
ICN [39]89.9777.7153.3873.2673.4665.0278.2290.7979.0584.8157.2062.1173.4570.2258.0872.45
ASBL-RetinaNet [65]89.5174.0746.9155.5473.7866.8778.4890.8670.0973.2046.7161.3470.5072.1732.8466.86
FENet (ours)88.4780.5454.6571.7078.0980.6587.3690.8184.5384.7453.2364.1476.8769.9457.6674.89