Table 1.

Evaluation indicators of a model’s predictive performance

Scenario	Name	Definition	Interpretation
Detection and segmentation	Recall	$R=\frac{\text{TP}}{\text{TP}+\text{FP}}$	The fraction of positive examples in the whole sample that are predicted to be correct
	Precision	$P=\frac{\text{TP}}{\text{TP}+\text{FN}}$	The fraction of true positive samples in the predicted positive samples
	Accuracy	$\text{ACC}=\frac{\text{TP}+\text{TN}}{\text{TP}+\text{FP}+\text{TN}+\text{FN}}$	The fraction of samples that are predicted to be correct out of all samples
	F-score	$F_1=\frac{2\text{TP}}{2\text{TP}+\text{FP}+\text{FN}}$	The harmonic mean of precision and recall
	Intersection over union	$\text{IoU}=\frac{P\cap G}{P\cup G}$	The fraction of the intersection of the predicted bounding boxes (P) and the ground-truth bounding boxes (G) to the union
Detection	Mean average precision	$\text{MAP}=\frac{1}{n}\sum _{k=1}^{k=n}{\text{AP}}_k$	The mean of the average scores of a group of queries [65], where n is the number of classes and APk is the average precision of class k
Segmentation	Dice coefficient	$\text{Dice}=\frac{2\text{TP}}{\text{FP}+2\text{TP}+\text{FN}}$	The function that evaluates the similarity of two contour regions

TP stands for true positive, FP stands for false positive, FN stands for false negative, and TN stands for true negative