Table 2 Performance comparison of different localization models for eyelid tumors
From: HM_ADET: a hybrid model for automatic detection of eyelid tumors based on photographic images
Model | Internal test set | External test set | ||||
|---|---|---|---|---|---|---|
Precision | Recall | AP | Precision | Recall | AP | |
YOLOv7 | 0.907 | 0.772 | 0.763 | 0.777 | 0.745 | 0.647 |
YOLOv7 + ConvNeXt | 0.918 | 0.795 | 0.796 | 0.847 | 0.724 | 0.692 |
YOLOv7 + ConvNeXt + FReLU | 0.878 | 0.850 | 0.836 | 0.825 | 0.782 | 0.745 |
YOLOv7_CNFG | 0.898 | 0.906 | 0.893 | 0.844 | 0.800 | 0.765 |