*3.2. Dataset*

The LS-SSDD-v1.0 dataset is widely used for SAR image intelligent interpretation [51–54]. The characteristic of small ships with large-scale backgrounds in LS-SSDD is close to actual satellite images; thus, we adopted the LS-SSDD-v1.0 dataset to verify the effectiveness of Lite-YOLOv5. Table 3 shows the details of the LS-SSDD dataset.

**Table 3.** Details of the LS-SSDD-v1.0 dataset.


From Table 3, there are 15 large-scale images (cover width ~250 km) numbered by 00.jpg to 15.jpg from different places (Tokyo, Adriatic Sea, etc.), polarizations (VV, VH), and scenes (land, sea). Considering the computing power of the GPU, we simply divided the original large-scale images into 800 × 800 sub-images without embellishment, keeping to the method of Zhang et al. [24]. Since there were fifteen 24,000 pixels × 16,000 pixels largescale images, the total sub-image number was 9000. Finally, according to Zhang et al. [24], the LS-SSDD-v1.0 dataset was divided into a training set for training learning and a test set for result performance evaluation via the ratio of 2:1.
