weizhiwang/OBELICS_HQ_5M_UniFilter
收藏官方服务:
资源简介:
UniFilter合成训练数据集包含大规模的多模态数据示例和质量分数对,用于训练统一的多模态数据质量分类器UniFilter。该数据集通过半合成方法生成,包括原始图像和对应文本,涵盖四个不同的质量等级。
The UniFilter Synthetic Training Data contains large-scale multimodal data examples and quality score pairs for training the Unified Multimodal Data Quality Classifier, UniFilter. The dataset is generated through a semi-synthetic approach, including raw images and corresponding text across four different quality levels.
提供机构:
weizhiwang


