foragi/try-v3
收藏资源简介:
该数据集是一个多媒体问答数据集,包含视频、音频和文本数据,旨在支持视频理解和问答任务。数据集由多个分片组成,每个分片对应不同的任务类型,如PR_correction(校正任务)、PR_event_reminder(事件提醒任务)、PR_post_event_reminder(后事件提醒任务)、RTP_world_knowledge(世界知识任务)、RTP_counting(计数任务)、RTP_fine_grained_movement(细粒度运动任务)、RTP_interaction_relation(交互关系任务)、RTP_OCR(光学字符识别任务)和RTP_Omni(全方位任务)。每个样本包括id、问题文本、两个答案、两个提醒、视频类型、视频时长、视频文件和问题音频文件。数据集总大小约为946MB,包含45个样本(每个分片5个样本),适用于训练和评估AI模型在多媒体场景下的理解和推理能力。
This dataset is a multimedia question-answering dataset containing video, audio, and text data, designed to support video understanding and QA tasks. It consists of multiple splits, each corresponding to different task types, such as PR_correction, PR_event_reminder, PR_post_event_reminder, RTP_world_knowledge, RTP_counting, RTP_fine_grained_movement, RTP_interaction_relation, RTP_OCR, and RTP_Omni. Each sample includes id, question_text, two answers, two reminders, video_type, video_duration, video file, and question_audio file. The total dataset size is approximately 946MB, with 45 examples (5 per split), suitable for training and evaluating AI models in multimedia scenarios for comprehension and reasoning.




