Statistical Adaptation Loss Improved SMALL Sample Ship Detection Method Based on an Attention Mechanism and Data Enhancement

Synthetic aperture radar (SAR) imagery is a promising data source for ocean activity detection. Ship target detection based on SAR images widely used in maritime trade and the military. image are rare, amount of public small. For applications ship detection, model with low dependence, fast iteration training cost needed. In this paper, balanced positive negative enhancement method was used. Through statistical analysis dataset, similar sea areas set filled targets comfortable size features. Increasing proportion samples helps to improve effect. The regional attention preadaptation mechanism implemented extract information, scale-adaptive loss combined accuracy model. Using same data, our exhibited better performance. When using 30% stable terms average precision (AP) maintained results achieved 100% dataset.