Likes 0Comments 0
Robot AI · training data · NVIDIA GEAR · 2026-10-25
NVIDIA publishes three Flora models for annotating robotics video
Three key points
- Flora comprises three models for video segmentation, captioning and outcome-reward labels.
- NVIDIA says its own engine annotates 200,000 hours a day across 512 GB300 GPUs.
- The figure is company-reported video annotation throughput, not new robot experience or independent validation.
Story summary
Robot Newspaper reported that NVIDIA GEAR released three Flora models for segmenting actions in video and adding captions and outcome-reward labels. NVIDIA's project page says its own production data engine labels 200,000 hours per day using 512 GB300 GPUs. This is a company-reported annotation throughput, not newly collected robot-experience hours or an independently verified performance metric.
Login required. Log in to recommend stories and write comments.
Comments 0
Be the first to comment.