Object Detection on PKU-DAVIS-SOD 1.0 (test)
52.5mAP@50CMTFormer
Evaluation Results
| Method | Links | |
|---|---|---|
| CMTFormerModality=Events + RGB, Input representation=Event image + RGB frame, Backbone=Deformable DETR, Temporal=Yes, Spatial-Prior=Yes2026.06 | 52.5 | |
| SODFormerModality=Events + RGB, Input representation=Event image + RGB frame, Backbone=Deformable DETR, Temporal=Yes, Spatial-Prior=No2026.06 | 50.4 | |
| Baseline (Deformable DETR)Modality=RGB, Input representation=RGB frame, Backbone=Deformable DETR, Temporal=Yes, Spatial-Prior=Yes2026.06 | 48.9 | |
| Deformable DETRModality=RGB, Input representation=RGB frame, Backbone=DETR, Temporal=No, Spatial-Prior=No2026.06 | 46.1 | |
| LSTM-SSDModality=RGB, Input representation=RGB frame, Backbone=SSD, Temporal=Yes, Spatial-Prior=No2026.06 | 45.6 | |
| Faster R-CNNModality=RGB, Input representation=RGB frame, Backbone=R-CNN, Temporal=No, Spatial-Prior=No2026.06 | 44.3 | |
| JDFModality=Events + RGB, Input representation=Channel image + RGB frame, Backbone=YOLOv3, Temporal=No, Spatial-Prior=No2026.06 | 44.2 | |
| MFEPDModality=Events + RGB, Input representation=Event image + RGB frame, Backbone=YOLOv3, Temporal=No, Spatial-Prior=No2026.06 | 43.8 | |
| YOLOv3-RGBModality=RGB, Input representation=RGB frame, Backbone=YOLOv3, Temporal=No, Spatial-Prior=No2026.06 | 42.6 | |
| Baseline (Deformable DETR)Modality=Event, Input representation=Event image, Backbone=Deformable DETR, Temporal=Yes, Spatial-Prior=Yes2026.06 | 33.4 | |
| Deformable DETRModality=Event, Input representation=Event image, Backbone=DETR, Temporal=No, Spatial-Prior=No2026.06 | 30.7 | |
| ASTMNetModality=Event, Input representation=Event embedding, Backbone=Rec-Conv-SSD, Temporal=Yes, Spatial-Prior=No2026.06 | 29.1 | |
| LSTM-SSDModality=Event, Input representation=Event image, Backbone=SSD, Temporal=Yes, Spatial-Prior=No2026.06 | 27.3 | |
| Faster R-CNNModality=Event, Input representation=Event image, Backbone=R-CNN, Temporal=No, Spatial-Prior=No2026.06 | 25.1 | |
| YOLOv3-RGBModality=Event, Input representation=Reconstructed image, Backbone=YOLOv3, Temporal=No, Spatial-Prior=No2026.06 | 24.4 | |
| NGA-eventsModality=Event, Input representation=Voxel grid, Backbone=YOLOv3, Temporal=No, Spatial-Prior=No2026.06 | 23.2 | |
| SSD-eventsModality=Event, Input representation=Event image, Backbone=SSD, Temporal=No, Spatial-Prior=No2026.06 | 22.1 |