Object Detection on M3FD (Class APs, mAP50, mAP)
95AP (Car)Semantic Bridge Fusion
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| Semantic Bridge FusionBackbone=CSPDarknet, Modality=RGB+IR2026.04 | 95 | 94.6 | 89.7 | 91.6 | 63.5 | |
| YOLOv8lBackbone=CSPDarknet, Modality=RGB2026.04 | 91.2 | 92.9 | 86 | 80.9 | 52.5 | |
| Semantic Bridge FusionBackbone=ResNet50, Modality=RGB+IR2026.04 | 91 | 92.7 | 87.7 | 84.3 | 55.8 | |
| YOLOv5lBackbone=CSPDarknet, Modality=RGB2026.04 | 90.5 | 91.2 | 86.1 | 82.7 | 52.5 | |
| YOLOv8lBackbone=CSPDarknet, Modality=IR2026.04 | 90 | 90.9 | 85.9 | 79.5 | 53.1 | |
| Faster-RCNNBackbone=ResNet50, Modality=RGB2026.04 | 87.9 | 81.3 | 72.4 | 74.5 | 42.3 | |
| TiNetBackbone=ResNet50, Modality=RGB+IR2026.04 | 87.5 | 83.1 | 81.6 | 79.8 | 49.5 | |
| YOLOv5lBackbone=CSPDarknet, Modality=IR2026.04 | 87.2 | 86.8 | 82.5 | 80.3 | 50.7 | |
| RI-YOLOBackbone=CSPDarknet, Modality=RGB+IR2026.04 | 83.6 | 56.6 | 91.6 | 88.5 | 47.8 | |
| RetinaNetBackbone=ResNet50, Modality=RGB2026.04 | 83.2 | 75.8 | 69.9 | 64.7 | 37.4 | |
| Faster-RCNNBackbone=ResNet50, Modality=IR2026.04 | 81.3 | 66.7 | 63.3 | 63.3 | 35.6 | |
| RetinaNetBackbone=ResNet50, Modality=IR2026.04 | 79 | 68 | 60 | 57.9 | 33.8 |