Dataset · RekaAI
First-person daily videos with captions and Q&A
RekaAI/RekaDaily-10k-processed
Millions of short daily first-person videos with detailed captions and Q&A from Reka AI.
What's inside
The dataset contains between 1 and 10 million entries clipped from non-gaming daily first-person videos. The videos were recorded by paid Claru platform participants at home and workplaces using head-mounted cameras and smartphones across various regions. Each English clip features a detailed text description and a series of second-person Q&A pairs.
What you can do with it
- SFT fine-tuning of multimodal video models for human action understanding (video-text-to-text)
- Training visual question answering (VQA) systems for wearable devices and smart glasses
- Tasks
- Video text → text, Visual question answering, Video classification
- Languages
- English
- Size
- 1M–10M rows
- Downloads this month
- 44K
- Likes
- 12
How to load it
from datasets import load_dataset
ds = load_dataset("RekaAI/RekaDaily-10k-processed")