Intelligent Safety Environment at LG Electronics Pyeongtaek Digital Park
LG Electronics Pyeongtaek Digital Park utilizes EVA to ensure worker safety and
provide rapid response to dangerous situations.
Traditional AI CCTV systems had limitations in the scenarios they could detect and struggled to adapt to new risks.
EVA enables a flexible and scalable safety response system by interpreting field requirements in natural language and autonomously expanding detection scenarios.
EVA x Cisco: Intelligent Safety Environment
By integrating Cisco Meraki cameras with EVA, a smarter and more responsive safety environment is established.
Safety personnel can configure detection scenarios using natural language and activate them instantly.
This enables rapid response to emerging risk situations across the site.
Scenario Refinement and Enrichment
| Case | Step 1 (Initial) | Step 2 (Describing False Alert Conditions) | Step 3 (Structured by Enrich Agent) |
|---|---|---|---|
| Fall Detection | Alert me when someone is lying on the ground. | Except for people standing or sitting on chairs, alert me when someone is lying on the ground | Current Case Detect people who have fallen on the ground Detection Steps - Person detected - At least one person is lying on the ground Exceptions - Person is sitting or standing - Hard to confirm person’s posture - Difficult to distinguish lying posture - Body occlusion > 50% - Objects covering the person or interfering with detection |
| Mask-Wearing Detection | Alert me when someone is not wearing a mask. | Among people sitting and working, alert me for those not wearing masks excluding those using laptops or smartphones | Current Case Detect workers not wearing masks while sitting and performing tasks Detection Steps - Sitting on a chair - Performing a task (action recognition) - Mask not worn on the face Exceptions - Hard to determine work activity - Face is covered or obstructed |
Basic rule settings alone can generate false alarms in diverse field situations.
With additional user feedback, EVA progressively enriches and refines scenarios.
EVA’s Enrich Agent structures such information to help AI understand real situations more precisely.
Feedback to Improve False Alarms
When a false alarm occurs, user feedback is stored in the Multi-modal Vector DB,
enabling EVA to continuously learn and improve performance based on site-specific characteristics.
Performance Enhanced by Operators
Operational teams can refine detection scenarios and resolve false alarms directly,
helping EVA evolve into an intelligence optimized for real-world field environments.
Performance further improves when combined with state-of-the-art foundation model updates.
Efficient Scalability
EVA integrates with Meraki camera infrastructure effortlessly,
delivering cost-efficient and scalable AI-driven safety operations across the cloud.
Flexible and Real-Time Safety Response System
EVA enables proactive detection of hazardous worker actions and helps prevent accidents before they happen.
With an evolving AI that reflects field feedback and adapts to dynamic conditions,
both operational efficiency and response speed continue to improve.
