SA-1B(segment anything)

分割任何10亿 (SA-1B) 是一个数据集，旨在训练来自开放世界图像的通用对象分割模型。 SA-1B 由 1100 万张多样化、高分辨率、隐私保护图像和使用我们的数据引擎收集的 1.1B 高质量分割掩码组成。它旨在用于我们的数据许可允许的计算机视觉研究。 这些图像是从一家大型图片公司获得许可的。1.1B 掩码是使用我们的数据引擎生成的，所有这些都是由 Segment Anything Model (SAM) 完全自动生成的。Segment Anything 10 Billion (SA-1B) is a dataset designed to train general-purpose object segmentation models from open-world images. SA-1B consists of 11 million diverse, high-resolution, privacy-protected images and 1.1 billion high-quality segmentation masks collected using our data engine. It is intended for computer vision research permitted under our data license. These images are licensed from a major stock photo company. The 1.1 billion masks were generated using our data engine, and all of them were fully automatically produced by the Segment Anything Model (SAM).