DramaQA for Video Story Understanding


Updates

2020-06-29 DramaQA Dataset v1.1 was released.

2020-05-07 DramaQA Dataset Paper was uploaded to Arxiv.

2020-04-09 DramaQA Challenge Leaderboard is open and Starter Code is released.

2020-04-09 DramaQA Dataset v1.0 was released.

2020-02-20 The 2nd Video Turing Test (VTT) : Toward Human-Level Video Story Understanding Workshop will be held in ECCV 2020, Glasgow, Scotland Online.

2019-11-11 DramaQA Challenge Leaderboard was open and Starter Code is released.

2019-11-02 The 1st Video Turing Test (VTT) : Toward Human-Level Video Story Understanding Workshop was held in ICCV 2019, Seoul, South Korea.

2019-10-31 The 1st DramaQA Challenge was launched.

2019-10-30 DramaQA Dataset v0.9 was released.





DramaQA for Video Story Understanding


Video Question and Answering (Video QA) is an emerging benchmark task to measure the levels of machine intelligence for video understanding.
While several previous studies have suggested datasets for the Video QA task, they did not consider story-level understanding, resulting in a lack of variance in the degree of question difficulty and coherent story-centric description.
DramaQA is a large-scale video QA dataset based on a Korean popular TV show, “Another Miss Oh”. This dataset contains four levels of QA on difficulty and multi-level character-centered story descriptions. We are expecting this dataset could be a starting point to evaluate human level video story understanding.




Papers


DramaQA: Character-Centered Video Story Understanding with Hierarchical QA

Seongho Choi, Kyoung-Woon On, Yu-Jung Heo, Ahjeong Seo, Youwon Jang, Seungchan Lee, Minsu Lee, Byoung-Tak Zhang

[PDF] [Code]