This repository provides videos and annotation json files of UFVideo-Bench, which including three tasks: PixRQA (integrating general QA, video object referring, and video segmentation), as well as PixHQA and PixTRQA (joint general QA, video object referring, video segmentation and temporal video grounding).
PixRQA, PixHQA and PixTRQA correspond to task1_bench, task2_bench and task3_bench respectively. Each JSON file example contains the relative path of… See the full description on the dataset page:
https://huggingface.co/datasets/Hevven/UFVideo-Bench.
