基于CenterNet的改进遥感旋转目标检测

刘鑫,黄进,杨瑛玮,李剑波

Improved Remote Sensing Rotating Object Detection based on CenterNet

Xin LIU,Jin HUANG,Yingwei YANG,Jianbo LI

表2 不同模型在UCAS-AOD数据集上精度评估结果

Table 2 Accuracy evaluation results of different models on the UCAS-AOD data set

Models Backbone AnchorFree Car(%) Plane(%) mAP@50(%) InputSize 两阶段算法 Faster RCNN[3] ResNet50 × 86.87 89.86 88.36 800 x 800 RoI Tran[14] ResNet50 × 88.02 90.02 89.02 800 x 800 RIDet-Q[34] ResNet50 × 88.50 89.96 89.23 800 x 800 单阶段算法 R-Yolov3[6] Darknet53 × 74.63 89.52 82.08 800 x 800 R-RetinaNet[31] ResNet50 × 84.64 90.51 87.57 800 x 800 VR-CenterNet DLAs √ 86.31 94.39 90.35 640 x 640