Table 5 Detailed per-class Average Precision (AP, in %) and overall mAP@0.5 on the GC10-DET dataset. The best performance for each metric is highlighted in bold.
From: Balancing complexity and accuracy for defect detection on filters with an improved RT-DETR
Model | Pu | Wl | Cg | Ws | Os | Ss | In | Rp | Cr | Wf | mAP@0.5(%) |
|---|---|---|---|---|---|---|---|---|---|---|---|
RT-DETR (baseline) | 94.26 | 92.40 | 94.70 | 81.28 | 57.50 | 45.02 | 23.61 | 27.30 | 58.70 | 64.10 | 65.48 |
EC-YOLO23 | 95.00 | 38.00 | 96.00 | 84.00 | 79.00 | 62.00 | 82.00 | 70.00 | 38.00 | 63.00 | 71.00 |
RT-DETR-LGA (ours) | 95.56 | 93.10 | 92.22 | 82.27 | 64.86 | 60.30 | 37.81 | 58.60 | 57.50 | 81.31 | 73.40 |
