Facebook open sources its DeepFocus VR research, which aims to create ultra-realistic visuals for eye-tracking headsets
including its code and the data set we used to train it—so the wider VR community can accelerate their research. http://twitter.com/... @fb_engineering : Facebook Reality Labs has built DeepFocus, a new AI-powered framework for rendering natural, realistic focus effects in real time. We're now open-sourcing our networks and data set. http://code.fb.com/... http://twitter.com/... Hugo Barra / @hbarra : Last #F8, @marifes gave us a peek at Half Dome, a prototype headset that showcases our research on varifocal displays. Today, learn more about DeepFocus, a new AI-powered rendering system that creates realistic, real-time retinal blur for a true-to-life visual experience in VR http://twitter.com/...
Context & Ripple Effects
DeepFocus is the rendering half of a bet Facebook laid out earlier this year, when CTO Mike Schroepfer framed AR expertise as what keeps the company relevant in 10 or 20 years; the system powers natural focus blur for varifocal displays like the [[a:entity|Half Dome]] prototype Hugo Barra demoed at F8.
The open-source move follows an established playbook: Facebook gave away its $30K Surround 360 camera designs in 2016, and later released the AI Habitat platform and Replica dataset on GitHub — each time seeding a research ecosystem with artifacts expensive to reproduce.
First-order effects
- VR researchers get Facebook Reality Labs' trained networks and dataset for free, removing the compute and data cost of reproducing real-time focus rendering and making Half Dome-style varifocal work easier to replicate.
Second-order effects
- Competing headset makers and academic labs now benchmark their eye-tracking rendering against Facebook's released baseline, pulling the research community toward Facebook's approach and its talent orbit.
Third-order effects
- If the Surround 360–DeepFocus–Habitat pattern holds, major labs will keep giving away trained models and datasets to set de facto standards in nascent hardware categories while keeping the product roadmap — headsets, glasses — proprietary.
The trend: Large corporate labs are open-sourcing expensive-to-train AI artifacts to shape emerging XR ecosystems, converting research generosity into standards-setting power.