Table 3 Comparison of benchmark evaluation results for target detection and recognition.
From: A Large-Scale Synthetic Benchmark Dataset for Non-Cooperative Space Target Perception
Method | mAP | Flops(G) | Params(M) | backbone | mAP_small | mAP_medium | mAP_large |
|---|---|---|---|---|---|---|---|
Faster R-CNN | 87.0 | 208 | 41.4 | Resnet50 | 75.8 | 85.4 | 91.5 |
YOLOv3 | 72.6 | 11.6 | 61.6 | Darknet53 | 60.4 | 69.9 | 78.7 |
Centernet | 80.8 | 12.1 | 32.1 | Resnet50 | 65.2 | 77.5 | 87.3 |
DETR | 78.3 | 96.5 | 41.6 | Resnet50 | 49.6 | 76.6 | 86.7 |
Sparse R-CNN | 93.7 | 152 | 106 | Resnet50 | 84.3 | 92.6 | 97.3 |
YOLOF | 87.3 | 99.0 | 42.6 | Resnet50 | 77.1 | 85.8 | 91.0 |
Deformable DETR | 78.0 | 193 | 40.1 | Resnet50 | 47.9 | 77.3 | 84.4 |
YOLOX | 84.5 | 13.3 | 8.94 | CSPdarknet | 72.3 | 82.4 | 89.2 |
ViTDet | 84.7 | 279 | 101 | ViT-base | 70.9 | 83.0 | 89.5 |
DiffusionDet | 88.5 | 105 | 111 | Resnet50 | 75.0 | 86.6 | 93.3 |
