From: Performance analysis of different DCNN models in remote sensing image object detection
Method | Backbone | mAP@.5 | mAP@[.5:.95] | Time |
---|---|---|---|---|
Faster R-CNN | ResNet50 | 0.750 | 0.497 | 96.5 ms |
RetinaNet | ResNet50 | 0.816 | 0.591 | 22.6 ms |
CenterNet | Harglass52 | 0.771 | 0.557 | 14.2 ms |
YoloV3 | DarkNet53 | 0.768 | 0.539 | 13.4 ms |
YoloV4 | CSPDarknet53 | 0.882 | 0.639 | 12.3 ms |
Our | CSPDarknet53* | 0.908 | 0.667 | 15.4 ms |