Memories-ai/UGC-VideoCap
收藏Hugging Face2025-10-05 更新2025-11-01 收录
下载链接：
https://hf-mirror.com/datasets/Memories-ai/UGC-VideoCap
下载链接
链接失效反馈官方服务：
资源简介：
UGC-VideoCaptioner数据集是一个针对短形式用户生成视频的详细全模态字幕生成的新型基准数据集。它包含了1000个TikTok视频，通过结构化的三阶段人工在环标注流程，覆盖了音频、视觉以及音频视觉联合的语义。数据集还提供了4000个精心设计的问答对，用于评估模型在单模态和跨模态理解方面的性能。
The UGC-VideoCaptioner dataset is a new benchmark for detailed multimodal captioning of short-form user-generated videos. It consists of 1000 TikTok videos annotated through a structured three-stage human-in-the-loop pipeline covering audio-only, visual-only, and joint audio-visual semantics. The dataset also includes 4000 carefully crafted QA pairs for evaluating the models performance in both unimodal and cross-modal understanding.
提供机构：
Memories-ai
