RRSIS-D (Referring Remote Sensing Image Segmentation – Dataset) is a large dataset specifically for remote sensing image segmentation tasks, launched by Xiamen University.Rotated Multi-Scale Interaction Network for Referring Remote Sensing Image Segmentation" has been accepted by CVPR 2024. The paper introduces the RRSIS task and the Rotational Multi-Scale Interaction Network (RMSIN) method.

The dataset contains 17,402 image-caption-mask triplets covering a variety of spatial resolutions and object orientations. It combines computer vision and natural language processing techniques to achieve pixel-level localization of target objects in remote sensing images based on text descriptions. It not only provides a wide range of spatial and rotational scenarios for the model, but also establishes a rigorous benchmark for the RRSIS task to ensure rigorous evaluation of performance.