davidfertube/hse-compliance-inspector

HSE Compliance Inspector

Vision-language model for automated Health, Safety & Environment (HSE) compliance inspection at industrial facilities.

Model Description

This model performs visual inspection of industrial worksites to detect HSE compliance violations including:

• PPE Compliance: Hard hats, safety glasses, high-visibility vests, steel-toe boots, gloves
• Housekeeping: Obstructed walkways, improper storage, spill hazards
• Fall Protection: Missing guardrails, unprotected edges, improper harness use
• Fire Safety: Blocked exits, missing extinguishers, improper storage of flammables
• Electrical Safety: Exposed wiring, missing lockout/tagout, damaged equipment

Architecture

Built on Qwen2-VL multimodal foundation with domain-specific fine-tuning on industrial safety imagery.

Input Image → Qwen2-VL Backbone → Safety Feature Extraction → Violation Classification → Report

Intended Use

• Pre-shift safety walkthroughs
• Automated audit documentation
• Continuous monitoring of work zones
• Training and compliance verification

Training Data

Trained on the hse-safety-inspection-data dataset containing labeled industrial safety inspection records.

Limitations

• Designed for industrial/energy facility environments
• Performance may vary in low-light or heavily occluded scenes
• Should supplement, not replace, qualified HSE personnel

Resources

• Demo: Vision Agent Space
• Dataset: hse-safety-inspection-data
• GitHub: vision-agent
• Portfolio: davidfernandez.dev

Built by David Fernandez — Applied AI Engineer | LangGraph & Pydantic Contributor