video_id stringclasses 3
values | segment_id stringclasses 3
values | subtitle_language stringclasses 3
values | subtitle_text stringclasses 3
values | action_label stringclasses 3
values | event_label stringclasses 3
values | fps int64 24 30 | resolution stringclasses 2
values | video_path stringclasses 3
values |
|---|---|---|---|---|---|---|---|---|
vid-001 | seg-001 | ps | خبروال د ورځې مهم خبرونه لولي. | news_reading | intro_segment | 25 | 1920x1080 | videos/segmented/vid-001-seg-001.mp4 |
vid-002 | seg-003 | en | Children enter the classroom and sit down. | classroom_activity | scene_transition | 30 | 1280x720 | videos/segmented/vid-002-seg-003.mp4 |
vid-003 | seg-002 | fa | مردم برای خرید به بازار می روند. | market_walk | public_activity | 24 | 1920x1080 | videos/segmented/vid-003-seg-002.mp4 |
ZamAI Pashto Video
ZamAI Pashto Video is a video understanding dataset scaffold for multilingual Afghan media research, with support for scene segmentation, subtitle alignment, and temporal event annotation.
Dataset Summary
The repository is structured for raw and segmented video assets, subtitle generation, action labels, and media metadata needed for temporal analysis workflows.
Languages
- Pashto
- Dari
- English
Modalities
- Video
- Audio
- Text
Project Structure
videos/raw/: Source video files.videos/segmented/: Scene-level or clip-level segments.videos/subtitles/: Subtitle files for supported languages.annotations/: Action labels, scene segments, and temporal events.metadata/: FPS, resolution, and domain metadata.scripts/: Frame extraction, scene segmentation, and subtitle generation.configs/: Dataset configuration.viewer/train.csv: Small sample table used by the Hugging Face dataset viewer.
Planned Fields
video_idsegment_idaction_labelsubtitle_languagefpsresolution
Intended Use
This repository supports multimodal video understanding, subtitle alignment, temporal event extraction, and scene-level Pashto media analysis.
Limitations
The current repository uses sample viewer rows for schema preview rather than full media assets. Video paths and subtitles in viewer/train.csv are representative examples intended to make the dataset viewer immediately useful.
License
This dataset repository is distributed under the Apache 2.0 license.
- Downloads last month
- 86