Table 4 Overall performance comparison on DDTE (vs. YOLOv11s: +2.7% \(\hbox {mAP}_{50}\), +209% FPS, −21.2% params).
Datasets | Methods | Params | Gflops | FPS | \(\hbox {mAP}_{50}\) | \(\hbox {mAP}_{50-95}\) |
|---|---|---|---|---|---|---|
DDTE | Faster R-CNN16 | 41.364 | 90.913 | - | 77.5 | 59 |
YOLOv5s | 9.13 | 192.67 | 65.74 | 80.4 | 57.8 | |
YOLOv8s | 11.14 | 229.5 | 65.15 | 78.8 | 56.9 | |
YOLOv10s | 8.07 | 198.33 | 64.35 | 77.5 | 55.4 | |
YOLOv11s | 10.73 | 213.06 | 62.56 | 80.4 | 58.2 | |
DAB-DETR17 | 44 | 216 | 120.35 | 79.6 | - | |
Deformable-DETR18 | 40 | 173 | 122.15 | 78.4 | - | |
DINO19 | 218 | - | 113.76 | 79.7 | - | |
LFF-YOLO20 | 60.51 | 6.85 | 168.49 | 78.9 | - | |
FFDDNet15 | 10.08 | 222.51 | 155.0 | 78.4 | 55.1 | |
LE-YOLOv521 | 4.8 | 10.3 | 180.0 | 79.9 | 57.1 | |
Ours | 8.46 | 159 | 193.5 | 82.6 | 61.6 |
