Atlas can turn a single reference image into novel views from any camera position and angle
XRarchitect · x · 2026-09-03
A new demo shows Atlas generating new views at any specified camera position and angle from just one reference image — a capability with clear implications for robot perception, simulation, and visual data augmentation. The author quips: "just a lonely robot."
Related event: Atlas Generates Novel Views From a Single Reference Image(2 posts)→
More from Embodied
- text-to-cad: open-source agent skills that turn plain language into CAD models and robot URDFs — tom_doerr · 2026-09-03
- motion-bricks.cpp hits GitHub: tiny model generates animations in realtime on a desktop CPU — zhengyiluo · 2026-09-03
- Robotics researcher corrects viral clip: end-to-end neural net robot video was teleop — chris_j_paxton · 2026-09-03
- NVIDIA details ROSCon 2026 lineup: Isaac Sim workshop and ROS 2 Lyrical memory transports — kscottz · 2026-09-03
- AgenticROS adds agent skills to robots with npm plugins and SKILL.md folders — chrismatthieu · 2026-09-03
- Cybercab night testing wraps at Giga Texas a day before Austin launch event — yunta_tsai · 2026-09-03