Thank you for using the AI Smart Safety Shield!
This Terms of Use agreement ("Terms") governs your installation, configuration, and operation of the AI Smart Safety Shield application, edge services, neural networks, and hardware interface assemblies (collectively, the "Services" or "System"). The System is designed as a hybrid millisecond-level proximity lockout safety net for operators working alongside industrial robotics and dangerous manufacturing machinery.
Please read these Terms carefully before deployment. By registering an account, booting up the neural edge tracker, or mounting the ESP32 relay modules, you agree to be bound by these Terms, including the safety guidelines and liability limitations referenced herein.
1. Scope of the System & Safety-First Directive
The AI Smart Safety Shield acts as a probabilistic software-driven barrier. It detects worker proximity through vision cameras and transmits electronic shutdown signals to high-speed relay contactors. You explicitly acknowledge and agree that:
- Not a Single Point of Safety: This software-based system must never replace required physical cage gates, physical light curtains, or hardwired mechanical e-stops required under OSHA, ISO 13849, or local industrial health directives.
- No Safety Bypasses: Under no circumstances should operators enter hazardous machinery cells without executing standard physical Lockout/Tagout (LOTO) protocols. Do not bypass mechanical guards under the assumption that the camera will halt the machine.
- Operator Awareness: Operators and maintenance workers bear ultimate responsibility for their personal safety and must visually verify machine power states prior to entry.
🚨 CRITICAL DANGER WARNING
Relying solely on visual AI models for protective lockouts poses a threat of severe injury or death. This system must be used in conjunction with standard physical machine guarding mechanisms.
2. Account Registration & Role Clearance Tiers
To access the System panels, you must secure an operator gateway account through Google Firebase Authentication. Account permissions are divided programmatically into three strict tiers. You agree to operate exclusively within your clearance limits:
- Viewer clearance: Read-only access to diagnostics, analytics charts, and telemetry. Viewers cannot modify camera states, map zones, or issue override commands.
- Operator clearance: Allowed to configure cameras, calibrate target coordinates, define zone margins, and toggle stream feeds. Operators must ensure camera lenses remain free of dust, grease, and obstacles.
- Admin clearance: Holds full administrative overrides. Admins can bypass zone lockouts and release emergency stops. The Admin tier bears full operational liability for override actions.
3. Technical Uptime & Stream Quality Duty
To operate effectively, the YOLOv8 proximity tracking model requires high-speed, stable network telemetry. Operating standard requirements include:
- Telemetry FPS & Latency: Feeds must stream at a minimum rate of 30 frames per second (640x480 resolution). Average transmission latency from camera sensor to AI microservice must stay below 100ms.
- Network Packet Loss: If network latency fluctuates beyond 150ms or packet loss exceeds 5%, the edge handler will immediately trigger a fail-safe shutdown.
- Relay Fail-Safe Configuration: All ESP32 contactor boards must be wired to open-collector default states, ensuring that power loss or serial disconnect results in a machine halt.
⚠️ NETWORK WARNING
Lapses in stream frequency directly delay the shutdown reaction time. Running camera feeds over low-bandwidth public links or congested Wi-Fi routes is strictly prohibited.
4. Data Acquisition, Privacy & Incident Logging
For audits and safety documentation, the System logs breach incidents in real-time:
- Violation Snapshots: When a worker breaches a warning or danger zone, the AI edge service captures a visual frame, encodes it as a base64 string, and saves it to a persistent MongoDB record.
- Automated SMTP Alerts: Captured snapshots are dispatched instantly via Mailcloud SMTP relays to operations directors and factory floor supervisors.
- Encryption Standards: Incident logs, machine coordinates, and account data are encrypted during transit and at rest using AES-256-CBC algorithms.
5. Limitation of Liability
The YOLOv8 neural network runs on probabilistic model architectures. Detection quality can be degraded by environmental elements (heavy dust, soot, ambient glare, smoke, safety clothes colors, camera line-of-sight blocks). Code Vortex Technologies Inc. provides this software "as is":
- No Guarantee: Code Vortex does not guarantee 100% detection accuracy under all conditions.
- Accident Liability Release: Code Vortex is not liable for accidents, injuries, machine wear, or factory downtime caused by sensor lag, cloud hosting latency (including Render cold-starts), local network failures, or custom ESP32 wiring issues.
- Liability Cap: In no event shall our total liability for any claim exceed the fees paid to Code Vortex Technologies for the software license in the preceding twelve-month period.