Enhanced YOLOv8 with DWR-DRB and SPD-Conv for Mechanical Wear Fault Diagnosis in Aero-Engines.

The results show that introducing the DWR-DRB module alone increased mAP50 from 85.4% to 87.0% ( p = 0.008, d = 1.12), indicating its significant effect on sparse feature extraction; the addition of SPD-Conv further improved mAP50 to 87.9% ( p = 0.012, d = 0.98), validating its role in preserving fine-grained features; the mAP50 of the complete model (integrating all modules) reached 91.0%, showing a highly significant difference compared to the baseline YOLOv8’s 85.4% ( p = 0.002, d = 2.05), with an effect size exceeding 1.5, proving that the performance improvement is practically meaningful.

also in 132,142 other papers