结合深度可分离卷积的多源遥感融合影像目标检测

陈江浩,杨军

Object detection for multi-source remote sensing fused images based on depthwise separable convolution

Jianghao CHEN,Jun YANG

表 3 本研究算法在VEDAI数据集上与其他主流算法的目标检测精度对比

Tab.3 Comparison of object detection accuracy between proposed algorithm and state of arts on VEDAI dataset

类别 AP50/% cars pickup camping truck tractor boat van 其他 YOLOrs[26] 83.48 76.96 65.69 53.51 69.07 22.28 56.88 43.88 YOLOfusion[27] 91.72 85.91 78.94 78.15 71.96 71.14 75.23 54.77 SuperYOLO[16] 91.61 86.80 79.25 89.33 86.39 54.26 81.51 68.79 MF-YOLO[17] 92.03 86.61 78.19 72.58 82.88 64.64 78.66 57.36 ICAfusion[21] 97.05 96.21 89.64 92.66 94.50 64.53 28.33 83.40 本研究算法 97.31 96.82 93.21 87.03 98.17 59.43 32.04 88.25