tferhan/Buildings_mask_morocco

该数据集包含超过100万个摩洛哥建筑物分割掩码，使用Meta AI开发的SAM3掩码生成模型生成。这些掩码是从高分辨率卫星图像（512×512瓦片，缩放级别17）推断出来的，覆盖约57,000个瓦片。所有输出都经过地理参考，可供下载、可视化和分析。每个掩码包括一个相关的置信度分数，以支持质量评估和下游任务的过滤。该数据集开放用于训练、研究和探索。

This dataset contains over 1 million building segmentation masks across Morocco, generated using the SAM3 mask generation model developed by Meta AI. The masks were inferred from high-resolution satellite imagery (512×512 tiles at zoom level 17), covering approximately 57,000 tiles. All outputs are georeferenced and ready for download, visualization, and analysis. Each mask includes an associated confidence score to support quality assessment and filtering for downstream tasks. This dataset is released for open use in training, research, and exploration.