AI Agent & Vision AI Technology

AI Agent analysis · separate solution
AI Agent AI Agent analysis · separate solution

The AI Agent is a separate solution that requires agreed data connections. It analyzes authorized data, summarizes results, exports CSV files, and recommends actions. Your team reviews the recommendations and carries them out.

Question-driven analysis

  • The AI Agent developed by P2ACH AI
  • Question-driven analysis

    The AI Agent is a separate solution that requires agreed data connections. It analyzes authorized data, summarizes results, exports CSV files, and recommends actions. Your team reviews the recommendations and carries them out.

P2ACH AI Agent dashboard visualization
LLM
Reasoning
Rule
Engine
Inference
Model
P2ACH AI Agent dashboard visualization mobile

Connected data analysis

Connected data analysis

P2ACH Ontology connects advertising, visits, paths, and shelf observations from all four solutions. With POS and sales data connected, the separate AI Agent can compare these authorized datasets using observed relationships and agreed business rules. It recommends sales and operating priorities for your team to review and act on.

  • Reasoning Engine

    Ask how visits, dwell, or campaign metrics changed over a selected period. The agent compares authorized connected data; sales analysis requires POS and sales data connections.

  • Recommendations

    The agent recommends priorities based on the observed data. Your team reviews the evidence and carries out campaign or store-operation changes.

Camera calibration for position and gaze estimates

01

Camera calibration for position and gaze estimates

  • Extrinsic Calibration
  • Extrinsic calibration is a camera calibration technology also used in autonomous driving.

    P2ACH AI applies it to retail spaces to correct distortion and field-of-view errors, then estimate the 3D position, direction, and coordinates of people or objects. That makes path and gaze analysis more accurate in physical spaces.

Applications
  • Advertising Effectiveness Analysis
    • Ad Viewership Analysis
    • Estimated attention to ad content
  • Customer Behavior Analysis
    • Frequently Gazed Products on Shelves
    • Heatmap Analysis

High Performance

02

High Performance

  • Object Detection & Tracking
  • Object Detection and Tracking demonstration
    Detection identifies people or objects, while tracking follows them in real time.

    When tracking is unstable, missed frames can cause the same person or object to be counted incorrectly. P2ACH AI reduces missed frames so ad performance, foot traffic, and heatmap analysis stay reliable.

    96.4%
    Detection performance validation results Detection performance validation results
    Technical tests and their measurement conditions

    Technical tests cover detection in advertising viewing areas and age-group and gender estimates. Review the reported results with their test conditions when assessing a deployment.

Applications
  • Advertising Effectiveness Analysis
    • Audience estimates: age, gender, and Re-ID
  • Customer Behavior Analysis
    • Path & Trajectory Tracking
  • Shelf gap detection
    • Check whether a selected product is visible in its assigned shelf position

Two camera inputs, one Edge AI Box

03

Two camera inputs, one Edge AI Box

  • Dual-Camera Processing
  • Dual CAM Processing with single Edge AI Box
    A single Edge AI Box can connect two cameras and run analysis models at the same time.

    Two-camera processing expands the analysis area and supports front-and-back measurement for double-sided devices without adding a second AI Box.

Applications
  • Wide Digital Signage Case
    • Wide Digital Signage dual-camera case

      Wide-Area Analysis

    • Dual-camera integration with a single device
    • Use two camera inputs to extend the measured area of a large display within their configured fields of view.
  • Double Side Digital Signage Case
    • Double-sided Digital Signage dual-camera case

      Double Side Angle Analysis

    • Dual-camera integration with a single device
    • Connect two cameras to one Edge AI Box to analyze both sides of a double-sided display. Measurement covers the configured camera views.

Movement and unique visitor analysis

04

Movement and unique visitor analysis

  • Re-Identification
  • Person Re-Identification multi-camera tracking Person Re-Identification multi-camera tracking mobile
    • Tracked shopper silhouette for Re-ID matching
    • Re-ID match overlay across camera views Re-ID match overlay across camera views mobile
    Re-ID matches observations of a person across configured camera views. It helps compare movement paths and reduce duplicate counts; it does not establish a person’s name or identity.
    Reduce duplicate counts
    Staff filters and Re-ID help reduce counting employees or repeat observations as new visitors.
    Tracking across cameras
    Compare paths across camera views where coverage and matching conditions support tracking.
    Compare exposure and subsequent movement
    Links ad exposure to subsequent actions, such as moving to specific product zones.
    Re-ID examples with multi-camera and staff recognition
    Re-identification: Examples of Multi-camera Re-ID & Staff Recognition.
Applications
  • Store operations
    • Customer movement across camera views
  • Advertising Effectiveness Analysis
    • Wide-Angle Ad Effectiveness Analysis via Multi-Camera Systems
    • Compare ad viewing, shelf visits, and purchases when POS data is connected
    • Staff Exclusion Feature and Unique Audience Data Calculation (Filtering Repeat Exposures)

Edge AI for supported devices

05

Edge AI for supported devices

  • Multi OS support
  • Multi-OS Edge AI Box compatibility
    Check hardware and OS requirements for your deployment
    OS compatibility
    We check OS, camera input, and processing requirements before selecting the deployment configuration.
    Android-based Edge AI Box
    The Android-based Edge AI Box supports AI processing and content playback on one device. We assess workload and compatibility for the planned setup.
    Customized Deployment
    We assess whether your existing workstation or content player meets the model and camera requirements.
Applications
  • Real-time Edge AI for Multi-Device Ad Insights
    • Powering AI models for ad effectiveness analysis on KOO's next-generation set-top boxes.
    • Deploying AI for ad performance and customer behavior analysis on tablet ordering systems.
    • Simultaneous content playback and AI model processing on our dedicated Edge AI Box.

Security controls for data processing and access

Security controls for data processing and access

LRQA approvals and ISO/IEC 27001 certification

GDPR and ISO27001 security certifications GDPR and ISO27001 security certifications mobile
    • GDPR (General Data Protection Regulation)

      A law enacted by the European Union

    • LRQA GDPR:2016 approval

      Applies to the service scope documented in the LRQA approval

    • ISO27001

      International standard for Information Security Management Systems (ISMS)

Homomorphic Encryption data security Homomorphic Encryption data security mobile
  • ISO/IEC 27001 certification badge ISO/IEC 27001 certification badge mobile
  • Data security covers processing, transmission, retention, and access. P2ACH AI uses encryption and access controls under the agreed deployment design. ISO/IEC 27001 certification and LRQA approval apply to their documented scopes.
Global market certifications for Korea Japan Europe US Global market certifications mobile

We are obtaining the necessary certifications to deliver our services across key global markets, including Korea, Japan, Europe, and the United States.

Data Anonymization process flow Data Anonymization process mobile

Data Anonymization for Ad Performance Analytics

Edge AI for supported devices Data Anonymization for Ad Performance Analytics

For audience reporting, on-device analysis produces aggregate metrics. Image retention, transmission, and access are defined separately for the selected solution and deployment.

  • ① Video data captured by external cameras is sent to an AI recognition device.
  • ② The Edge AI device processes camera input and derives audience metrics.
  • ③ Reporting metrics are transmitted under the agreed data flow. Any retained images or events follow separate retention and access rules.
  • ④ Anonymized data is processed by a statistics web server to deliver the service.