Staer releases synthetic warehouse dataset for physical-AI research
Staer released the first version of its synthetic warehouse dataset for robotics perception and navigation research.
Evidence notes
- The initial release covered 15 warehouse environments, 1,500 sequences and roughly 200,000 object instances.
- Each frame aligned RGB, depth, semantic-segmentation and instance-segmentation data generated in NVIDIA Isaac Sim.
- The rolling Hugging Face dataset has since advanced to v0.2 with stereo walkthroughs, IMU data and scene-graph annotations.