2.3 Model selected
Liu et al. (2021) used six indexes (Accuracy, Precision, Recall, F1 value, ROC curve, and AUC) to compare the gully recognition results and accuracy evaluation of the U-Net, R2U-Net, and SegNet image semantic segmentation models. The SegNet model ranked first for gully recognition in the hilly and gully region of the Loess Plateau, followed by the R2U-Net and U-Net models (Liu et al. , 2021). The gully length and width between predicted and measured values had RMSE values of 6.78 m and 0.50 m, respectively, using the SegNet model, indicating its superior performance for gully recognition and morphological feature extraction. Hence, this study used the SegNet model for gully recognition and morphological feature extraction at the watershed scale. Figure 2 is a network structure diagram of the SegNet model.