Research map / Vision-language and multimodal / Visual question answering
Towards Video Text Visual Question Answering: Benchmark and Baseline
Minyi Zhao, Bingjia Li, Jie Wang, Wanqing Li, Wenjing Zhou, Lan Zhang, Shijie Xuyang, Zhihang Yu, Xinkun Yu, Guangze Li, Aobotao Dai, Shuigeng Zhou
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add OCR features (text, visual, temporal) on top of video features (ablation) | T5-ViteVQA · M4-ViteVQA Task1Split1 val | accuracy: 11.51 -> 23.17 | Improved |
| add OCR visual features (ablation) | T5-ViteVQA · M4-ViteVQA Task1Split1 val | accuracy: 20.79 -> 23.17 | Improved |
| add OCR temporal features (start/end temporal ids) (ablation) | T5-ViteVQA · M4-ViteVQA Task1Split1 val | accuracy: 19.78 -> 23.17 | Improved |
| add video features Xv (ablation) | T5-ViteVQA · M4-ViteVQA Task1Split1 val | accuracy: 22.36 -> 23.17 | Improved |
| use T5-ViteVQA instead of M4C (temporal reasoning) | T5-ViteVQA vs M4C · M4-ViteVQA Task1Split1 val | accuracy: 17.91 -> 23.17 | Improved |
| use T5-ViteVQA instead of M4C on easy questions | T5-ViteVQA vs M4C · M4-ViteVQA Task1Split1 val | accuracy: 19.30 -> 25.09 | Improved |
| use T5-ViteVQA instead of M4C on hard questions | T5-ViteVQA vs M4C · M4-ViteVQA Task1Split1 val | accuracy: 9.02 -> 14.26 | Improved |
| use T5-ViteVQA instead of M4C on text questions | T5-ViteVQA vs M4C · M4-ViteVQA Task1Split1 val | accuracy: 17.26 -> 23.08 | Improved |
| use T5-ViteVQA instead of M4C on vision questions | T5-ViteVQA vs M4C · M4-ViteVQA Task1Split1 val | accuracy: 18.36 -> 24.21 | Improved |
| use T5-ViteVQA instead of M4C on harder split (cross-video-source) | T5-ViteVQA vs M4C · M4-ViteVQA Task1Split2 test | accuracy: 11.36 -> 16.68 | Improved |
| use T5-ViteVQA instead of M4C in domain adaptation Task2 without adaptation | T5-ViteVQA vs M4C · M4-ViteVQA Task2 test | accuracy: 7.52 -> 9.29 | Improved |
| use T5-ViteVQA instead of the VideoQA model JustAsk | T5-ViteVQA vs JuskAsk · M4-ViteVQA Task1Split1 val | accuracy: 10.81 -> 23.17 | Improved |
Closest papers by meaning
- Track the Answer: Extending TextVQA from Image to Video with Spatio-Temporal Clues (AAAI 2025 · similarity 0.93)
- Scene Text Visual Question Answering (ICCV 2019 · similarity 0.92)
- Structured Multimodal Attentions for TextVQA (TPAMI 2021 · similarity 0.91)
- When Open-Vocabulary Visual Question Answering Meets Causal Adapter: Benchmark and Approach (AAAI 2025 · similarity 0.91)
- VTQA: Visual Text Question Answering via Entity Alignment and Cross-Media Reasoning (CVPR 2024 · similarity 0.91)
- Not Another Text Benchmark: Putting the “Visual" Back in Visual Question Answering for Large Video Models (NeurIPS 2026 · similarity 0.90)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
