Dataset · jinlong17
CTVid-Bench video text restoration benchmark
jinlong17/CTVid-Bench
An ECCV 2026 test dataset for evaluating clear text restoration in video and VQA.
What's inside
An open benchmark from the ClearText-Video paper (accepted at ECCV 2026) combining video restoration and scene-text readability improvement tasks. The public release includes media files for three variants (GT original, blur, and 4x downsampling), as well as current QA v2 annotations in English and Chinese.
What you can do with it
- Evaluating video restoration and scene-text enhancement models
- Testing OCR and visual question answering quality on blurred or compressed frames
- Comparing video upscaling algorithms with small text
- Tasks
- Visual question answering, Video classification
- Languages
- Chinese, English
- Size
- 10K–100K rows
- Downloads this month
- 22.3K
- Likes
- 0
How to load it
from datasets import load_dataset
ds = load_dataset("jinlong17/CTVid-Bench")