Abstract: The current deep-learning based SAR ship instance segmentation models fail to consider the full level information of features and the context information of targets, which leads to low instance segmentation accuracy. In order to address this problem, a SAR ship instance segmentation method based on a full-level context information squeeze-and-excitation region of interest (ROI) extractor is proposed. This method proposes a novel ROI extractor(ROIE), called FL-CI-SE-ROIE. First of all, FL-CI-SE-ROIE can extract ROI at all levels, which can retain the full level features of the target, thus enhancing the multi-scale description capability of the network. Then, FL-CI-SE-ROIE expands the ROI context information, which can gain the context information of targets, thus enhancing the background identification capability. Finally, FL-CI-SE-ROIE introduces a squeeze and excitation(SE) module, which can balance ROI context information in different ranges, thus suppressing background interference, and further improving the accuracy of instance segmentation. The experimental results on the public polygon segmentation SAR ship detection dataset (PSeg-SSDD) show that the SAR ship instance segmentation accuracy of the proposed method is higher than that of the current 9 comparison models.