Table 2 Comparison of detection results with other methods.
From: SED-YOLO based multi-scale attention for small object detection in remote sensing
Model | YOLOv3- tiny | YOLOv5s | RetinaNet | ARSD41 | MDCNet42 | CoF-Net43 | LO-Det44 | Ours |
|---|---|---|---|---|---|---|---|---|
Plane | 42.1 | 67.5 | 74.6 | 86.9 | 92.7 | 89.7 | 89.2 | 70.0 |
Ship | 68.3 | 85.4 | 52.1 | 79.0 | 88.5 | 82.1 | 84.2 | 86.4 |
Storage-tank | 86.9 | 92.0 | 37.2 | 79.0 | 60.2 | 83.9 | 81.3 | 92.8 |
Baseball-diamond | 46.6 | 70.4 | 33.6 | 72.2 | 80.6 | 55.5 | 66.1 | 80.9 |
Tennis-court | 66.9 | 86.7 | 61.2 | 90.7 | 92.9 | 89.6 | 90.7 | 89.3 |
Basketball-diamond | 75.6 | 82.7 | 51.6 | 72.2 | 80.6 | 72.9 | 75.1 | 83.9 |
Ground-track field | 43.6 | 58.0 | 34.1 | 56.9 | 56.6 | 69.3 | 56.0 | 58.2 |
Harbor | 48.8 | 51.8 | 32.1 | 72.2 | 83.0 | 54.5 | 59.9 | 44.7 |
Brige | 90.7 | 92.8 | 21.2 | 46.4 | 48.5 | 41.6 | 31.3 | 93.3 |
Small vehicle | 51.4 | 60.5 | 38.4 | 80.6 | 65.5 | 65.1 | 65.1 | 62.2 |
Large vehicle | 63.3 | 74.9 | 29.7 | 67.0 | 86.5 | 77.2 | 71.0 | 77.1 |
Helicopter | 43.3 | 59.3 | 13.3 | 50.6 | 35.1 | 59.0 | 48.4 | 59.7 |
Roundabout | 45.2 | 60.9 | 36.8 | 61.3 | 59.3 | 66.1 | 59.3 | 58.3 |
Soccer-ball field | 32.4 | 47.0 | 29.4 | 79.0 | 59.8 | 72.8 | 44.7 | 50.9 |
Swimming pool | 42.3 | 54.9 | 38.2 | 72.4 | 68.8 | 58.8 | 54.9 | 66.4 |
mAP50(%) | 56.5 | 69.2 | 38.9 | 68.3 | 69.2 | 69.3 | 66.2 | 71.6 |
IT(ms) | 2.2 | 2.3 | 28.7 | 23.2 | — | 47.6 | 16.7 | 4.0 |
Params(M) | 8.7 | 7.0 | 36.4 | 13.1 | — | 37.6 | 26.9 | 8.0 |
GFLOPs | 12.9 | 15.9 | 213.6 | 68.0 | — | — | — | 35.3 |
