Object Detection on DSEC-Detection (mAP50)
50.6mAP50CMTFormer
Evaluation Results
| Method | Links | |
|---|---|---|
| CMTFormerModality=RGB +Events, Backbone=Deformable DETR, Head=Deformable DETR, Temporal=Yes, Spatial-Prior=Yes, #Params=87.9M2026.06 | 50.6 | |
| Our baseline*Modality=RGB, Backbone=Deformable DETR, Head=Deformable DETR, Temporal=Yes, Spatial-Prior=Yes, #Params=46.4M2026.06 | 49.4 | |
| Swin TransformerModality=RGB, Backbone=Swin-T, Head=Mask R-CNN, Temporal=No, Spatial-Prior=No, #Params=44.8M2026.06 | 48.7 | |
| SODFormerModality=RGB +Events, Backbone=Deformable DETR, Head=Deformable DETR, Temporal=Yes, Spatial-Prior=No, #Params=84.8M2026.06 | 47.8 | |
| ConvNeXtModality=RGB, Backbone=ConvNeXt-T, Head=Mask R-CNN, Temporal=No, Spatial-Prior=No, #Params=45.6M2026.06 | 46.2 | |
| Deformable DETRModality=RGB, Backbone=DETR, Head=DETR, Temporal=No, Spatial-Prior=No, #Params=39.4M2026.06 | 41.3 | |
| SFNetModality=RGB +Events, Backbone=CSPDarknet53, Head=YOLOv5, Temporal=No, Spatial-Prior=No2026.06 | 41.2 | |
| Faster R-CNNModality=RGB, Backbone=ResNet50+FPN, Head=Faster R-CNN, Temporal=No, Spatial-Prior=No, #Params=41.9M2026.06 | 35.4 | |
| CenterNetModality=RGB, Backbone=Hourglass-104, Head=CenterNet, Temporal=No, Spatial-Prior=No, #Params=191.4M2026.06 | 35.1 | |
| YOLOv5Modality=RGB, Backbone=CSPDarknet53, Head=YOLOv5, Temporal=No, Spatial-Prior=No, #Params=76.8M2026.06 | 33.2 | |
| YOLOv7Modality=RGB, Backbone=E-ELAN, Head=YOLOv7, Temporal=No, Spatial-Prior=No, #Params=151.7M2026.06 | 31.5 | |
| RetinaNetModality=RGB, Backbone=ResNeXt-101-FPN, Head=RetinalHead, Temporal=No, Spatial-Prior=No, #Params=53.9M2026.06 | 30.5 | |
| Our baseline*Modality=Event, Backbone=Deformable DETR, Head=Deformable DETR, Temporal=Yes, Spatial-Prior=Yes, #Params=46.4M2026.06 | 29.1 | |
| FPN-fusionModality=RGB +Events, Backbone=ResNet-50, Head=RetinalHead, Temporal=No, Spatial-Prior=No2026.06 | 28.9 | |
| Swin TransformerModality=Event, Backbone=Swin-T, Head=YOLOv3, Temporal=No, Spatial-Prior=No, #Params=44.8M2026.06 | 26.6 | |
| EMS-YOLOModality=Event, Backbone=EMS-ResNet10, Head=YOLOv3, Temporal=No, Spatial-Prior=No, #Params=6.2M2026.06 | 20.2 | |
| Deformable DETRModality=Event, Backbone=DETR, Head=DETR, Temporal=No, Spatial-Prior=No, #Params=39.4M2026.06 | 18.7 |