ClarusC64/no_go_zone_compliance_v01

ClarusC64/no_go_zone_compliance_v01

Dataset summary

This dataset checks whether models respect no-go zones inside a container.
A no-go zone is a region where entities must not appear for functional or safety reasons.
The dataset includes allowed zones, restricted zones, and no-go zones across multiple domains.

Core behavior tested

• detect when an entity enters a no-go zone
• classify zone status per frame
• maintain zone compliance under occlusion
• track transitions back to safe areas

Domains

• factory lines (maintenance areas)
• warehouses (pedestrian vs robot lanes)
• labs (biohazard containment)
• sports pitches (referee-only regions)
• airports (runway vs taxi lanes)
• hospitals (sterile fields)
• construction sites (heavy machinery areas)

Key columns

• zone_type: type of region inside the container
• zone_status: allowed, restricted, or no_go
• zone_violation: true if entity enters no_go
• target_expected_zone: where the entity should be
• drift_risk: likelihood of spatial failure

Sample failure questions

• does the model treat restricted areas as normal?
• does occlusion hide a violation?
• does motion prediction cross forbidden boundaries?

Example usage

from datasets import load_dataset

ds = load_dataset("ClarusC64/no_go_zone_compliance_v01")

frame = ds["eval"][0]

print(frame["zone_id"], frame["zone_status"], frame["zone_violation"])