Table 4 Performance comparison on the UCF-EL dataset.
From: Hyperbolic geometry enhanced feature filtering network for industrial anomaly detection
Method | F1 | w-F1 | Acc | mIoU |
|---|---|---|---|---|
Deeplab v3 | 0.690 | 0.950 | 95.41 | 57.30 |
UNET | 0.766 | 0.957 | 96.01 | 52.13 |
Deeplab v3+ | 0.774 | 0.958 | 96.02 | 52.12 |
PAN | 0.770 | 0.957 | 95.90 | 51.01 |
PSPNet | 0.762 | 0.954 | 95.71 | 49.44 |
FPN | 0.773 | 0.958 | 95.93 | 51.02 |
SegFormer | 0.820 | 0.962 | 96.21 | 56.53 |
Ours | 0.853 | 0.971 | 96.83 | 59.41 |
