Gemini Robotics ER 2 is being used to auto-annotate 69,000+ robot videos
DynamicWebPaige · x · 2026-08-04
Google DeepMind’s Gemini Robotics ER 2 is being used to auto-annotate robot videos
The post points to an experiment where Gemini Robotics ER 2 is used as an auto-annotator for LeRobot video footage. The idea is to watch robot demonstrations and generate richer annotations automatically.
The underlying problem is that many of the 69,000+ robot datasets on Hugging Face’s LeRobot are still weakly labeled: most episodes only have a single task string, with no subtasks or timestamps. That makes them hard to use for training VLA/VAM/WAM-style systems.
The author says the model seems to work “pretty well” and asks whether the tool should be open sourced.
More from Embodied
- Raspberry Pi caption device turns calls and room talk into live subtitles — tom_doerr · 2026-08-04
- Researchers teach a humanoid robot to learn from its own mistakes — imjustnewatai · 2026-08-04
- H3 video model reportedly fixes the gymnast-proportion problem and supports 3D motion reconstruction — andrew_n_carr · 2026-08-04
- Waymo CEO says its first self-driving demo took 18 months, product took 15 years — Y Combinator · 2026-08-04
- ScienceCorp launches Scifi 2 headstage for brain recording and stimulation — gottapatchemall · 2026-08-04
- FCC ban on new foreign-made advanced robots explained in a new video — carlosdponx · 2026-08-04