embodiedSSM/UrbanVideo-Bench
收藏资源简介:
该基准测试旨在评估视频-大语言模型(Video-LLMs)是否能像人类一样自然地处理连续的第一人称视觉观察,包括回忆、感知、推理和导航等能力。数据集包含两部分:5k+的多选题问答(MCQ)数据和1k+的视频片段。文本数据存储在MCQ.parquet文件中,视频数据存储在videos文件夹中。MCQ.parquet文件包含多个字段,如Question_id(每个多选题的全局索引)、video_id(每个多选题对应的视频名称)、question_category(每个多选题对应的任务类别)、question(问题和选项的文本)和answer(多选题的正确答案)。
The benchmark is designed to evaluate whether video-large language models (Video-LLMs) can naturally process continuous first-person visual observations like humans, enabling recall, perception, reasoning, and navigation. The dataset consists of two parts: 5k+ multiple-choice question-answering (MCQ) data and 1k+ video clips. The text data is available in the file `MCQ.parquet`, and the video data is stored in the `videos` folder. The `MCQ.parquet` contains the following fields: `Question_id` (Global index of each MCQ), `video_id` (Video name for each MCQ), `question_category` (Task category corresponding to each MCQ), `question` (Text of question and options), and `answer` (Ground truth answer for the MCQ).




