Skip to content

Dataset · nvidia

Synthetic video of warehouses and hospitals

nvidia/PhysicalAI-SmartSpaces

280+ hours of synchronized video from 1,800 cameras for tracking and 2D/3D detection.

What's inside

The dataset is synthetically generated in Omniverse and Cosmos Transfer, containing over 280 hours of video from nearly 1,800 cameras. Scenes include warehouses, hospitals, retail stores, and other smart spaces. The video is time-synchronized and annotated for tracking people, forklifts, pallet jacks, and autonomous mobile robots (AMRs), as well as for 2D/3D object detection. The modality is video, and the metadata size is 1K–10K.

What you can do with it

  • Training multi-camera tracking
  • Training indoor 2D/3D object detectors
  • Testing robots and AMRs in synthetic scenes
Size
1K–10K rows
Downloads this month
49.7K
Likes
108

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("nvidia/PhysicalAI-SmartSpaces")