Dataset · ayushsdev
Complete QVHighlights Benchmark Video Set
ayushsdev/qvhighlights-videos
25,124 English MP4 video clips for text-to-moment retrieval tasks.
What's inside
The dataset contains a complete archive of 25,124 video clips (totaling 135.3 GB) from the well-known QVHighlights benchmark (NeurIPS 2021), split into train, val, and test subsets. Files are in MP4 format and organized into subfolders by the first character of the filename to bypass Hugging Face directory file limits. Source clips are from YouTube and belong to the English-language segment.
What you can do with it
- Training multimodal models for moment retrieval (highlights) in videos via text descriptions
- Evaluating quality of video-text-to-text understanding systems on a standard academic benchmark
- Tasks
- Video classification, Video text → text
- Languages
- English
- Size
- 10K–100K rows
- Downloads this month
- 25.4K
- Likes
- 0
How to load it
from datasets import load_dataset
ds = load_dataset("ayushsdev/qvhighlights-videos")