Journal of Remote Sensing / 2021 / Article / Tab 2

Research Article

Feature Enhancement Network for Object Detection in Optical Remote Sensing Images

Table 2

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

MethodC1C2C3C4C5C6C7C8C9C10C11C12C13C14C15C16C17C18C19C20mAP

[19]54.074.563.380.744.872.560.075.662.376.076.846.457.271.868.353.881.159.543.181.265.1
[60]54.278.063.281.041.272.660.779.162.978.377.056.959.371.462.353.581.156.343.481.365.7
[59]53.976.663.280.940.272.560.476.362.576.075.946.557.471.868.353.781.062.343.081.065.2
54.278.363.381.046.772.661.680.166.478.476.757.259.671.665.553.881.258.843.381.266.5
CornerNet [58]58.884.272.080.846.475.364.381.676.379.579.526.160.637.670.745.284.057.143.075.964.9
Libra R-CNN [61]62.578.872.080.846.772.664.479.969.177.576.346.259.371.868.053.981.162.443.281.367.4
FENet (ours)54.178.271.681.046.579.065.276.569.679.182.252.057.671.971.862.381.261.243.381.268.3