Solutions / Intelligent Video Analytics for Car Dealerships (IVA)

The system is based on IVA (Intelligent Video Analytics) principles and HOI (Human-Object Interaction) technology. Unlike standard surveillance systems, our solution performs deep post-analysis of visitor and employee behavior using YOLO (v10/v11) neural networks. Data is processed in Batch Processing mode overnight, providing consolidated reporting for the previous day by the start of the next work shift.

Implementing intelligent video analytics helps to:

• Objectively assess the performance quality of each sales manager;
• Digitalize the showroom "touchpoint funnel": from entry to opening a specific car door;
• Identify car models that generate high interest but have low conversion rates to test drives;
• Optimize staff schedules based on actual floor load patterns;
• Eliminate the human factor in collecting attendance and interaction statistics.

How the Solution Works (Batch Cycle)

The solution minimizes the load on the dealership's IT infrastructure during business hours. The process is divided into three stages: collection, nightly processing, and morning reporting.

• Data Collection (Day): Local storage (NVR/Server) records high-quality video streams from cameras without active analysis.
• Processing (Night): Compressed video archives are processed on a dedicated GPU node. The neural network recognizes human skeletal models (Pose Estimation) and their interactions with vehicle elements (opening doors, hood).
• Analytics (Morning): The system generates reports, calculates employee KPIs, and visualizes data in a dashboard available to management before the workday begins.

Technological Superiority of the NVIDIA Stack

The choice of the NVIDIA stack (DeepStream SDK, TensorRT, CUDA) is driven by the need for rapid processing of massive video datasets within a short nightly window:

• High Throughput: TensorRT technology accelerates neural network inference by 5–10 times, allowing 100 hours of video archives to be processed in just a few hours.
• HOI Detection Accuracy: The use of Tensor Cores enables high-precision models for recognizing complex actions, such as entering a vehicle cabin.
• Cost-Efficiency: Maximum processing density per GPU reduces the total cost of ownership (TCO) compared to CPU-based solutions.

Business Metrics and KPIs

The system provides detailed analytics divided into staff performance indicators and exhibit popularity:

• Sales Department Efficiency:
  • Response Time: time elapsed before a manager approached a client at a vehicle;
  • Consultation Duration: actual duration of interaction at the exhibit;
  • Staff Coverage: percentage of visits handled by staff according to regulations.
• Vehicle Interaction:
  • Model Interest: time spent by clients within a 1.5m radius of the car;
  • Car Body Interaction: count of door (front/rear), trunk, and hood openings;
  • Interior Engagement: time spent by the client inside the cabin.
• General Analytics: unique traffic (Footfall), Heatmaps, and showroom path analysis.

Technical Requirements and Components

To ensure proper system operation and high analytical accuracy (95%+), the following technical parameters must be met:

• Camera Requirements:
  • Resolution: at least 1080p (Full HD);
  • Frame Rate: 15–25 FPS;
  • Protocols: RTSP / ONVIF;
  • Placement: 3–4 meters high with viewing angles that minimize vehicle occlusions.
• Computing Resources:
  • Locally: a server with storage for the daily archive (H.264/H.265);
  • For Analysis: NVIDIA GPU (RTX 3060/4060 for Edge solutions) or L4 instances in the cloud.
• Network: Outbound internet speed must allow the transfer of the compressed archive to the cloud during the night (if cloud processing is selected).
• Privacy: Data processing can be performed entirely within the Customer's closed loop. External reports contain only anonymized digital metrics.

Business Process Integration

Video analytics results integrate with existing dealership systems to build end-to-end reports:

• CRM Integration: correlating visit times and showroom activities with transaction data.
• ERP/HR Systems: automated data export for calculating manager bonuses based on traffic handling quality.
• BI Platforms: data export via API for visualization in third-party systems.

Mobile Client

The mobile client (Android, iOS) displays key indicators and charts for companies using Sanatel analytics. The app provides managers with quick access to core metrics derived from the company's video analytics. Data security is ensured as only aggregated daily data is transmitted to the app, without granular details.