To meet the needs of sea surface target detection based on Unmanned Surface Vessel (USV), this paper designs the sea surface target detection system including the process of data collection and transmission. Hardware selection, interface connection and transmission protocol are also clarified. Based on the USV target detection system, a visible light sea target detection dataset (STDetection) is collected for detection and classification, including 3115 images, with cutter, trawler, ship, drum, buoy, and pier category. The statistics of the distribution of the target occupying the ratio of the image and target size in the image are illustrated. We tested the performance of commonly adopted algorithms on the dataset, such as Faster-RCNN, Cascade-RCNN, SSD and YOLO series. This dataset provides a benchmark of sea target detection and classification for evaluating the performance of detection algorithm. The dataset is released at https://github.com/echo-sky/STDetection.