DAMO-NLP-SG/VL3-Syn7M
收藏Hugging Face2025-02-07 更新2025-02-15 收录
下载链接：
https://hf-mirror.com/datasets/DAMO-NLP-SG/VL3-Syn7M
下载链接
链接失效反馈官方服务：
资源简介：
re-caption数据集是用于VideoLLaMA3模型训练的数据集，包含700万张具有短描述和详细描述的多样化高质量图片。这些图片来源于COYO-700M、MS-COCO 2017、CC-3M和LLaVA-Pretrain，并通过InternVL2进行了重新注释。数据集用于视频理解任务。
The re-caption dataset is used for training the VideoLLaMA3 model, consisting of 7 million diverse, high-quality images each accompanied by a short and detailed caption. The images are sourced from COYO-700M, MS-COCO 2017, CC-3M, and LLaVA-Pretrain, and have been re-annotated using InternVL2 for video understanding tasks.
提供机构：
DAMO-NLP-SG
