PhysicalAI SDG-Warehouse: Synthetic Industrial Safety Video Scenes

PhysicalAI SDG-Warehouse is a synthetic video dataset of staged industrial-safety events in a simulated warehouse environment. It contains approximately 123,000 video clips totaling roughly 412 hours of footage at 1920x1080 resolution and 30 frames per second. The dataset was created by NVIDIA and last updated on HuggingFace in May 2026.

Use Cases

• Training anomaly detection models for warehouse safety based on staged events like forklift near-misses.
• Developing and benchmarking video-based action recognition systems for industrial operations.
• Generating synthetic training data for computer vision models to recognize safety hazards.
• Simulating and studying evacuation procedures based on the warehouse fire scenario.

Strengths

• Approximately 123,000 video clips provide a substantial volume of training data.
• Footage is fully annotated and captured at a consistent 1920x1080 resolution and 30 fps.
• Data is organized across four distinct, staged industrial-safety scenarios.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Data is synthetic and generated in a simulated environment, which may not perfectly reflect real-world conditions.

Provenance

Source

NVIDIA

Collection Method

Synthetically generated and fully annotated in a simulated warehouse environment.

Freshness

Last updated 2026-05-29 22:49:17; freshness should be verified.