Skip to content

Dataset · nvidia

NVIDIA synthetic warehouse operations video dataset

nvidia/PhysicalAI-WorldModel-Synthetic-Warehouse-Operations-Scenes

About 123,000 fully annotated synthetic video clips covering warehouse safety scenarios.

What's inside

The dataset contains approximately 123,000 video clips (about 412 hours) at 1920x1080 resolution and 30 frames per second with English annotations. The clips are generated in a simulated warehouse environment and cover four scenarios: unsafe forklift-worker proximity, fire with evacuation, forklift collision with shelving, and standard operation. The video file format in the card is not specified.

What you can do with it

  • Training video analytics systems for industrial safety control
  • Training and evaluating video classification and video-to-text models
  • Training text-to-video generative models and physical world simulators (World Models)
Tasks
Video classification, Video text → text, Text → video
Languages
English
Size
100K–1M rows
Downloads this month
48.5K
Likes
35

How to load it

from datasets import load_dataset

ds = load_dataset("nvidia/PhysicalAI-WorldModel-Synthetic-Warehouse-Operations-Scenes")