Inst-IT/Inst-It-Dataset
收藏Hugging Face2025-03-01 更新2025-04-12 收录
下载链接：
https://hf-mirror.com/datasets/Inst-IT/Inst-It-Dataset
下载链接
链接失效反馈官方服务：
资源简介：
Inst-IT数据集是一个包含21k个视频和51k个图像的大规模指令微调数据集，专为多模态实例理解任务设计。它提供了针对每个实例的细粒度注释，包括视频级描述、帧级描述、实例级描述以及开放式问答对。数据集通过显式的视觉提示指令微调来提升模型对多模态实例的理解能力。
The Inst-IT Dataset is a large-scale instruction-tuning dataset containing 21k videos and 51k images, designed for multimodal instance understanding tasks. It provides fine-grained annotations for each instance, including video-level descriptions, frame-level descriptions, instance-level descriptions, and open-ended QA pairs. The dataset enhances models understanding of multimodal instances through explicit visual prompt instruction tuning.
提供机构：
Inst-IT
