Dataset · fpvlabs
Stera-10m: egocentric multimodal dataset
fpvlabs/stera-10m
200 hours of synchronized first-person recordings with RGB, LiDAR depth, audio, and text for embodied AI and robotics.
What's inside
An open egocentric multimodal dataset for embodied AI, robotics, world models, and spatial intelligence. Collected on an iPhone Pro via the Stera platform, it features 200 hours of synchronized first-person recordings, over 500 sessions, 20 participants, 20+ environments, 10 million RGB frames, and LiDAR depth, alongside audio, video, 3D, and text. The language is English, and the metadata volume is 1 to 10 million.
What you can do with it
- Train policies for robots and embodied agents
- Fine-tune video and depth models
- Evaluate spatial intelligence and world models
- Tasks
- Robotics, Video classification, Image → text, Depth estimation
- Languages
- English
- Size
- 1M–10M rows
- Downloads this month
- 72.2K
- Likes
- 88
- Access
- on request on Hugging Face
How to load it
from datasets import load_dataset
ds = load_dataset("fpvlabs/stera-10m")