LLaVA CC3M
收藏arXiv2025-09-30 收录
下载链接:
https://huggingface.co/datasets/liuhaotian/LLaVA-CC3M-Pretrain-595K
下载链接
链接失效反馈官方服务:
资源简介:
该数据集是从CC3M数据集中筛选出的59.5万个图像的子集,用于对模型进行预训练。此外,该数据集与LLaVA-v1.5模型结合使用,以提升模型的性能。其任务是针对视觉-语言模型的预训练。
This dataset is a subset of 595,000 images filtered from the CC3M dataset, designed for model pre-training. Additionally, it is utilized in combination with the LLaVA-v1.5 model to improve the model's performance. The target task of this dataset is the pre-training of vision-language models.
提供机构:
OpenAI



