Stera: Open-Source Infra That Turns iPhones into Spatial Data for World Models
FPV Labs has launched Project Stera, an open-source pipeline that transforms consumer iPhones into high-fidelity spatial data capture devices for embodied AI research. The release includes the Stera-10M dataset, comprising over 200 hours of data and 10 million frames collected using the system. By providing accessible, standardized tooling, Stera aims to democratize access to multimodal real-world data and reduce fragmentation in AI data collection.
- ▪Project Stera is an open-source, end-to-end pipeline that enables iPhones to capture high-fidelity spatial data for AI training.
- ▪The Stera-10M dataset contains over 200 hours of data and more than 10 million frames captured using the Stera system.
- ▪Stera captures multimodal data including RGB, depth, IMU, 6DoF poses, hand movements, and semantic task information.
- ▪The Stera SDK allows users to read, process, and export data for evaluation and model training.
- ▪Stera aims to eliminate reliance on proprietary hardware like Aria and enable widespread, standardized data collection for embodied AI.
Opening excerpt (first ~120 words) tap to expand
We are releasing Project Stera - an open source, end-to-end pipeline that turns a commodity iPhone into a research-grade capture system for embodied AI training data.Today, we're open-sourcing the whole stack, along with Stera-10M, a 200+ hour dataset, and 10M+ frames captured entirely through it.FPV Labs began with one bet - the scaling law for embodied AI will need high-fidelity, multimodal real-world data, and the underlying infrastructure that produces this at scale without compromising downstream quality will determine how fast we build a general-purpose model.Over the last 12 months, we've seen how high-fidelity data is locked behind gated hardware like Aria, which is out of reach for researchers, builders, and startups that want to work with high-quality multi-modal data, and how…
Excerpt limited to ~120 words for fair-use compliance. The full article is at Ycombinator.