The Ultimate Guide to People Counting Technology: 3D, Edge AI Cameras, and WiFi Compared

TL;DR: 

• People counting is now essential for safety, smart building automation, retail optimization, transportation planning, and urban mobility insights.

• Four main technologies:

  • WiFi counting: cheapest but low accuracy (<50%), useful only for rough outdoor footfall trends.

  • 3D depth sensors: highest accuracy (~99%), ideal for entrances but costly with limited field of view.

  • 2D cameras + Edge AI: best balance of accuracy (95–98%), privacy, scalability, and cost; supports zone counting and queue monitoring.

  • 2D cameras + Cloud AI: advanced analytics (heatmaps, journeys) but high bandwidth, compute cost, and privacy burden.

• Most organizations today migrate toward 2D Edge AI because it offers the lowest total cost of ownership and strong accuracy without heavy infrastructure or privacy risks.

  
  

Introduction: Why People Counting Matters More Than Ever

In today's operational landscape, accurate people counting data has become an indispensable asset for organizations across every sector. From ensuring public safety to driving profitable business decisions, real-time occupancy management is now a core requirement.

1. Buildings & Safety: Beyond Basic Capacity

   Real-time occupancy monitoring is no longer optional—it's a safety requirement.

   Many building regulations mandate that all occupants must be able to evacuate within 8 minutes during emergencies such as fires.

   Because of this, facility managers rely on accurate, real-time people counting to ensure safe and compliant operations.

   • Smart Building Operations: Attendance data enables the automation of smart building systems, adjusting lighting, HVAC (heating, ventilation, and air-conditioning) based on actual usage, leading to significant energy savings.

   • Corporate Canteens: Canteens optimize operations by using attendance data to accurately plan food portions, thereby reducing food waste. Displaying average waiting times helps stagger employee arrivals, saving time and improving the efficiency of the Return-to-Office strategy.

   • Entertainment Venues: These spaces require enhanced safety protocols, especially for vulnerable groups, ensuring a secure environment for elderly patrons, children, and individuals with disabilities.

     
     

2. Retailers: Optimizing Conversions and Customer Experience

   Retail footfall counting is crucial for understanding attendance trends, which informs better decision-making regarding staffing levels, targeted promotions, and overall customer experience. Long waiting times at checkout are a major factor in customer dissatisfaction and reduced return rates.

   • Commercial Potential Evaluation: Foot traffic data is widely used by advertisers and prospective tenants to evaluate a location's commercial viability, as high pedestrian traffic often correlates with better visibility and higher conversion opportunities.

     
     

3. Transportation, Airports, and Cities

   • Transportation Operators: They rely on passenger counting data to appropriately size their services, ensuring comfortable rides while maintaining profitable operations.

   • Cities and Urban Planning: Cities need pedestrian mobility insights to reorganize urban landscapes, optimize public spaces, and determine retail pricing or attract new brands to city centers.

The Evolution of People Counting Solutions

The landscape of people counting technologies has matured significantly. Today, the dominant solutions fall into four main categories: (1) Infrared sensors, (2) WiFi-based counting, (3) 3D Depth Sensors (Stereo/ToF), and (4) 2D Camera-based systems using either Edge AI or Cloud AI.

Note: For this comparison, we exclude LiDAR solutions due to their considerably higher cost compared to the technologies discussed.

1. WiFi Beacon / Probe-Based People Counting

   How it works

   WiFi gateways detect the (typically anonymized or hashed) MAC addresses of nearby smartphones. Algorithms then estimate the number of unique devices.

   Strengths

   • Low Cost: Often leverages existing WiFi infrastructure.

   • Wide Range: Gateways can cover up to 100 meters.

   Limitations

   • Low Accuracy (Below 50%): Challenges include multiple devices per person, disabled WiFi, and MAC address randomization, which severely impacts reliability.

   • No Directionality: Cannot accurately distinguish between people entering and leaving.

   Ideal Use Cases

   Suitable only for very rough foot traffic analysis or for cities with limited budgets requiring basic path mapping and extrapolated estimates. This solution is quickly becoming obsolete due to privacy features like MAC randomization.

2. 3D Camera-Based People Counters (Stereo Cameras or 2D Camera + Time-of-Flight Sensor)

   How it works

   3D people counting devices detect the height and shape of individuals using either stereo vision or a combination of a 2D camera and a Time-of-Flight (ToF) sensor. They are typically installed above doorways or in corridors, capturing movement from a bird’s-eye perspective.

   Strengths

   • Highest Accuracy (Close to 99%+): The ability to measure height and depth virtually eliminates errors from shadows or overlapping bodies.

   • Maximized Privacy: The top-down view naturally avoids capturing identifiable facial features, ensuring inherent data privacy compliance.

   Limitations

   • Highest Hardware Cost: Due to the dual-camera or specialized ToF components.

   • Limited Field of View: Can only count vertically within a restricted area, making it unsuitable for wide or open spaces without multiplying units.

   Ideal Use Cases

   The preferred solution for smart buildings, precise occupancy management, and highly accurate retail entrance counting (conversion rates). Also widely deployed in airports and public transit hubs where absolute counting accuracy is essential for operations and security.

3. 2D Camera-Based People Counting: Edge AI vs. Cloud AI

   This category utilizes standard 2D cameras paired with Artificial Intelligence for detection and counting.

• 2D Cameras + Edge AI

  How it works

  Processing occurs directly on the camera device using specialized AI chips. The embedded model detects and counts people as they cross a virtual line, without continuous video transmission.

  Strengths

  • High Privacy: Only metadata (the count, not the video) leaves the device.

  • Low Bandwidth: Analysis happens on-device, drastically reducing network load.

  • Real-Time & Scalable: Provides immediate insights and scales easily across numerous distributed locations.

  • Accuracy: Offers a strong balance with accuracy generally between 95% and 98%.

  Limitations

  • Limited Processing Power: Edge AI may struggle in extremely crowded or highly occluded environments where people significantly overlap.

  • Dependence on Setup: Accuracy relies heavily on camera quality, positioning, and mounting angle.

  Ideal Use Cases

  Well-suited for city mobility studies, retailers with wide entrances seeking the best accuracy-to-cost ratio, and queue monitoring in canteens or retail, where high accuracy (95-98%) is sufficient.

  
  

• 2D Cameras + Server/Cloud AI

  What it is

  By processing full video rather than compressed metadata, server or cloud AI can support more advanced analytics, including heatmaps, journey paths, and behavioral insights.

  Standard CCTV or IP cameras stream full video to an external server or cloud platform, where powerful AI models perform the analysis. This architecture allows for integration with existing camera infrastructure.

  Strengths

  • Flexibility & Advanced Analytics: Powerful server compute supports more sophisticated features like heatmaps, journey paths, and detailed behavioral insights.

  Disadvantages

  • High Operational Cost: Continuous video streaming demands high bandwidth and substantial server computing, increasing processing and storage expenses.

  • Major Privacy Concerns: Full video footage is streamed and often stored, making GDPR/CCPA compliance significantly more complex and resource-intensive.

  Ideal Use Cases

  Suitable for environments with established CCTV systems (e.g., shopping malls, stadiums, airports) that require broad video analytics capabilities beyond simple counting.

Comparative Summary Table

Technology Type	Accuracy	Cost	Privacy Level	Installation Complexity	Scalability	Ideal Scenarios	Key Limitations
WiFi / Probe-based Counting	Low Under 50%	Low	Very high (no images)	Very easy	High	Outdoor footfall, city analytics, malls, events	Dependent on smartphone signals, MAC randomization, no directionality
3D Depth Sensors (ToF / Stereo / Structured Light)	Very High (99%+)	Medium–High	Very High (bird eye view)	Medium (needs overhead mounting)	Medium	Retail entrances,buildings, transport, security	Higher hardware price, limited counting zone, installation height constraints
2D Cameras + Edge AI	Medium–High (90%-98%)	Medium	High (metadata only leaves device)	Medium (gives more flexibility and options than 3D counters)	High	Retail, offices, real-time occupancy, existing camera replacement	Accuracy drops in crowded / occluded scenes; depends on camera angle/quality
2D Cameras + Server/Cloud AI	High (95%–98% assuming better models)	High ( bandwidth and compute)	Medium (video leaves device)	Medium (gives more flexibility and options than 3D counters)	Medium	Malls, large venues, CCTV-enabled buildings	Privacy compliance burden; high bandwidth; cloud compute can be costly
Infrared Beam Counters (legacy)	Low–Medium (60–85%)	Low	Very high	Easy	High	Small shops, basic entrance counting	No occlusion handling, miscounts in groups, no analytics

How to Choose the Right People Counting Technology

Based on your priorities, the most suitable technology can be selected:

1. If Accuracy is Primary (Especially at Entrances): 3D Depth Sensors remain the most reliable option (e.g., Xovis, V-Count, Acorel, and Milesight).

2. If Cost, Privacy, and Scalability are Key: The industry is rapidly shifting toward 2D Cameras with Edge AI. This technology offers a superior balance of accuracy (95-98%), low operational cost, and high data privacy, making it the preferred successor to older WiFi and Infrared systems.

3. Crucial Consideration: Only 2D camera-based approaches (Edge/Cloud) inherently support multi-zone counting and advanced queue monitoring. Due to the limited field of view of 3D sensors, they become economically inefficient when trying to cover large, open areas.

FAQs

Q1: How does MAC address randomization affect WiFi people counting?

MAC address randomization, a privacy feature in modern smartphones, causes the device to report a temporary, fake MAC address. This drastically reduces the accuracy of WiFi counting because the system cannot reliably identify and track unique, returning visitors, resulting in accuracy often dropping below 50%.

Q2: Is 3D people counting better than 2D Edge AI for retail?

3D sensors offer marginal superior accuracy (up to 99%+) and are ideal for narrow retail entrances where conversion rates are a key KPI. However, 2D Edge AI is often preferred for wide entrances, monitoring internal store zones, or for queue management due to its better cost-per-square-meter coverage and higher flexibility.

Q3: Which solution provides the lowest operational cost?

2D Edge AI typically provides the lowest total cost of ownership (TCO). While the hardware cost is medium, its operational costs are extremely low because it requires minimal bandwidth (only metadata is sent) and eliminates expensive cloud computing charges, unlike Server/Cloud AI solutions.