Green100K

该数据集是为了训练DiffuMatting模型而收集的大规模绿幕数据集，旨在生成高质量的 matte-level 标注。该数据集不仅用于合成一般物体和肖像抠图集合，而且是DiffuMatting模型训练过程中的关键。其规模属于大型，任务包括在纯绿色背景上生成物体以及进行matte-level的标注任务。

This dataset is a large-scale green screen collection curated for training the DiffuMatting model, with the objective of generating high-quality matte-level annotations. It not only supports the synthesis of general object and portrait matting datasets, but also serves as a critical component in the training pipeline of the DiffuMatting model. Boasting a large scale, its core tasks include generating objects on pure green backgrounds and completing matte-level annotation tasks.