Automated segmentation and identification of ships in satellite imagery is of utmost importance mainly due to its several crucial applications like maritime surveillance, fishing surveillance causing loss of millions of dollars to the nations, etc. But segmentation in such scenarios is a tough task due to the tiny size of the ships in the satellite images, thereby, keeping it an open area of research. Hence, the paper proposes a novel ship segmentation model for Satellite Aperture Radar (SAR) imagery by leveraging High-resolution network model with the integration of various attention mechanisms namely, self-attention, multi-head attention and cross modal attention, with cross-modal attention providing the best architecture. The work also conducts an investigation and analysis of various deep learning models including UNet [8], SegNet, HRNet, etc., for baseline model selection and comparative analysis. The results prove that the proposed model performed the best as compared to the other models with the highest Dice-coefficient of 78.75% and PSNR of 21.27%.