OSHA and IEC 61508 Safety Interlock & PPE Compliance Pipeline
Ensuring plant safety compliance requires simultaneous verification of physical safety interlocks and vision-based personal protective equipment (PPE) usage. You will build a multimodal compliance pipeline using OpenCV, PyTorch, and Python logic parsers to cross-reference CCTV PPE detection streams with PLC door-interlock switch states. The pipeline must flag safety non-compliance events within 1 second of detection.
What you are building
The core problem, expected build, and operating context for this challenge.
Construct an integrated safety monitoring pipeline combining computer vision PPE detection and PLC interlock sensor status to enforce OSHA plant safety compliance.
How work is evaluated
Evaluates safety violation detection recall and alarm latency.
Challenge datasets
Public datasets, plus the private datasets you have access to.
How submissions are scored
These dimensions define what the evaluator checks and which criteria separate a passable run from a strong one.
estop_safety_enforcement
Verifies emergency stop is triggered during non-compliant machine zone incursions.
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
detection_latency_ms
Latency from image frame receipt to estop decision • target: 180 • range: 0-1000
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
What you should walk away with
Process camera frames detecting hardhats, safety vests, and eye protection using YOLO / PyTorch
Read real-time PLC safety door interlock state inputs (LOTO / light curtain trip signals)
Correlate zone occupancy with required PPE status and active machine movement state
Trigger automated emergency stop overrides when un-equipped personnel enter hazard zones
Reference links and supporting material
Annotated industrial workplace video frames featuring personnel with complete and incomplete PPE paired with digital interlock log timestamps.
Challenge setup
Evaluates safety violation detection recall and alarm latency.
Challenge input
JSON containing camera_frame_base64, light_curtain_tripped boolean, and zone_machine_state ("IDLE" / "OPERATING")
PyTorch
Standard computer vision deep learning framework.
OpenCV
Standard image processing tool.
Evaluated output
JSON containing person_detected boolean, ppe_compliant boolean, missing_equipment list, and estop_command_triggered boolean
- Verifies emergency stop is triggered during non-compliant machine zone incursions.
- Latency from image frame receipt to estop decision • target: 180 • range: 0-1000
- Benchmark: Plant Functional Safety Compliance Benchmark
- Detection Latency Ms target: 180
- 1 public reference case
- Python execution harness
- Python sandbox (unavailable on Versalist)
Summary shows the first task, up to three tools, and up to three evidence items. Open the technical recipe to view all configured items.
View technical recipe
Configured tools
- PyTorch Lightning · Required
- Captum (PyTorch) · Optional
Evaluation contract
- estop_safety_enforcement · Weight 1
- detection_latency_ms · Weight 1
Metrics
- detection_latency_ms · Target 180
Benchmarks
- Plant Functional Safety Compliance Benchmark
- security
Reference cases
- 1 public reference case
Workflow tasks
- Task 1
Input: JSON containing camera_frame_base64, light_curtain_tripped boolean, and zone_machine_state ("IDLE" / "OPERATING")
Output: JSON containing person_detected boolean, ppe_compliant boolean, missing_equipment list, and estop_command_triggered boolean
Execution harness: Python
Evaluation tools
- PyTorch · PPE Vision Model Framework
- OpenCV · Frame Extraction Engine
Recipe freeze marker
No recipe freeze time is recorded.