NVIDIA open-sources Flora, a model suite that auto-labels robot videos for training GR00T
qinzytech · x · 2026-10-28
NVIDIA's GEAR lab released Flora, an open model suite for scaling robotics data engines. It turns robot and human manipulation videos into training labels: when each action happens, what it is, and how well it went.
Components:
- Flora-TempSeg splits a video into individual actions.
- Flora-Captioner writes a timestamped instruction for each action.
- Flora-Rewarder judges an episode: success or not, where it slipped, and why.
Inference code is on GitHub with weights on Hugging Face (no training code). NVIDIA says it uses Flora to train GR00T; anyone can clone the repo to caption and score their own robot or human videos.
Related event: NVIDIA Open-Sources Flora, a Robot Data Engine(2 posts)→
More from Embodied
- Hemlock launches EMG wearables to capture force data robots can't get from video — ycombinator · 2026-10-28
- Stringman: open-source $795 cable robot covers a whole room, with MCP server — GregKamradt · 2026-10-28
- Dyna Robotics aims to put 100 million robots in the world, hiring full-stack talent — JasonMa2020 · 2026-10-28
- From Vietnam soccer kid to Dyna robotics scientist after a breakout training feature — JasonMa2020 · 2026-10-28
- Mitosis Bench: AI agents find only 32.4% of dividing cells in tumor images — yawnxyz · 2026-10-28
- Building an interactive ESP32-S3 badge with Codex — fully offline — OpenAIDevs · 2026-10-28