*3.4. Evaluation Indices*

Precision (*P*) is calculated by

$$P = \frac{\#TP}{\#TP + \#FP} \times 100\% \tag{17}$$

where # denotes the number, *TP* denotes the situation where the prediction and label are both ships, and *FP* denotes the situation where the prediction is a ship but the label is the background.

Recall (*R*) is calculated by

$$R = \frac{\#TP}{\#TP + \#FN} \times 100\% \tag{18}$$

where *FN* denotes the situation where the prediction is the background but the label is a ship.

The average precision (*AP*) is calculated by

$$AP = \int\_0^1 P(\mathcal{R}) \cdot d\mathcal{R} \tag{19}$$

where *P* denotes the precision, and *R* denotes the recall.

*F*1 can take account of both precision and recall and is calculated by

$$F1 = 2 \times \frac{R \times P}{R + P} \tag{20}$$

Finally, *t* denotes the inference consuming time of a sub-image detection. As a result, the running time *T* of a large-scale image is equal to 600 *t*. Moreover, in order to evaluate the portability performance of Lite-YOLOv5 on on-board SAR ship detection, we also calculated the parameter size, FLOPs, and model volume.
