GRD: An Ultra-Lightweight SAR Ship Detector Based on Global Relationship Distillation

Most existing works on lightweight synthetic aperture radar (SAR) ship detectors sacrifice a lot of detection accuracy to reduce model size. In this letter, we propose an ultra-lightweight detector based on distillation technology, which can reduce the parameter quantity of the model while minimizing the damage to the model’s detection accuracy. Due to the scattering interference and speckle noise in SAR images, directly applying the existing ultra-lightweight detectors cannot achieve satisfactory performance for ship detection. As a result, we design a global relationship distillation (GRD) algorithm for the ultra-lightweight SAR ship detector. This algorithm can preserve more global relationships from the teacher and mitigate the accuracy degradation caused by the noise and interference, especially in complex inshore scenarios. Besides, the features learned by this algorithm are robust, and the pruned model is more stable. The superiority of the GRD method over several state-of-the-art distillation methods has been evaluated on the high-resolution SAR images dataset (HRSID).

I. Introduction

Synthetic aperture radar (SAR) [1] ship detection is an essential branch of SAR image interpretation. For ship detection tasks in low latency requirements scenarios such as maritime traffic safety, the time, and communication bandwidth costs of transmitting SAR data back to the ground station for processing are too high. Therefore, directly interpreting SAR images on unmanned aircraft or satellite processors will significantly improve efficiency. However, the storage space and computing power of these unmanned platforms are limited, which cannot meet the deployment of deep neural networks with large parameter quantities. Therefore, researchers have begun to explore models with lighter and lower power [3].