Table 2 Ablation results for WTConv and PFPN.

From: Efficient target detection method based on wavelet transform and progressive feature pyramid network: a case study of power grid inspection

Model

WTConv

PFPN

Precision

Recall

mAP50

mAP50-95

Parameters (M)

FPS

Baseline

✗

✗

0.918

0.792

0.841

0.62

5.97

24.1

+WTConv

✓

✗

0.941

0.808

0.859

0.631

5.45

26

+PFPN

✗

✓

0.933

0.804

0.854

0.628

5.76

25

WTConv + PFPN (full)

✓

✓

0.962

0.81

0.872

0.64

4.83

28.5

Back to article page