AIPro Provides AI People Counting at 63 Building Rooftop

For safety and facility managers overseeing large mixed-use complexes, high-rise towers, and public infrastructure, real-time occupancy management and data-driven preventative safety are top priorities. Especially in high-risk zones like observation decks, rooftops, and terraces, knowing the exact headcount inside a space at any given moment has become a core requirement of modern facility safety regulations.

Yet operations teams routinely hit a frustrating roadblock. While they need accurate entry and exit data, retrofitting an established building that has been operating for decades with new overhead 3D sensors or dedicated counting cameras means dealing with steep cabling costs, structural alterations, and disruptive operational downtime.

What if you could achieve 98% accurate foot-traffic counting and visitor analytics using your building's existing CCTV cameras—without installing a single piece of new hardware?

Here is how a recent deployment at the iconic 63 Building, carried out in technical partnership with the Advanced Security Cooperative, sets a new benchmark for smart safety management in large-scale facilities.

Accurately Tracking Rooftop Access Without Construction or Privacy Headaches

As one of Korea’s premier landmark high-rises, the 63 Building sees heavy daily foot traffic from both visitors and on-site staff. Its rooftop access doors are critical control points where accurate, real-time headcount tracking is essential for facility security and safety compliance. However, conventional people-counting solutions on the market fell short of the site’s operational requirements:

Hardware retrofits and infrastructure disruption: Traditional infrared (IR) beams and dedicated stereo cameras require drilling into doorframes to mount new units and running fresh network cabling. In a landmark building, anything that compromises architectural aesthetics or requires heavy construction is a non-starter.
Challenging lighting and erratic foot traffic: Doorways linking indoor spaces to an open rooftop experience harsh backlighting and drastic lighting shifts throughout the day. On top of that, people often walk through side-by-side or pause in the doorway to chat—causing basic line-crossing sensors to frequently double-count or miss entries altogether.
Strict privacy compliance requirements: Recording and storing high-resolution footage of visitors' faces and identifiable features on external servers creates severe data privacy liabilities and audit burdens under strict privacy regulations.

Ultimately, the 63 Building needed a software-first solution—one that leverages existing security cameras, maintains pinpoint accuracy in tough lighting and crowded conditions, and eliminates privacy risks completely.

Overcoming Site Challenges with 17-Point Pose Estimation and De-Identified Analytics

Step 1. Zero camera replacements—solving occlusion and crowd overlap with 17-point pose estimation Standard security cameras are typically mounted at angled perspectives, making occlusion—where one person blocks another from view—a constant challenge. To solve this, the system uses 17-point pose estimation technology.

Rather than relying on basic rectangular bounding boxes, the AI tracks 17 key skeletal joints—including the head, shoulders, elbows, hips, and ankles—in real time to map each person's 3D posture. Even when groups walk closely together, harsh sunlight creates glare, or visitors pass through pushing strollers or carrying bags, the engine isolates every individual trajectory and counts them with 98% accuracy.

Step 2. Beyond raw headcounts—multi-dimensional demographic and behavioral insights AIPro’s people-counting platform goes far beyond logging "+1 in, -1 out." Its AI engine classifies visitor gender, age group (children, adults, seniors), and specific movement patterns in real time. Facility teams can spot peak hours for children or seniors to optimize safety staff rotations, or detect unusual behavior—such as loitering or falls near access doors—to trigger immediate safety responses and inform smarter building operations.

Step 3. Eliminating security and privacy risks with a non-storage, de-identified engine To meet the uncompromising security standards of a major commercial high-rise, the entire analytics pipeline runs on a de-identified statistics engine. Live video streams from existing CCTV feeds are processed in-memory, converted instantly into anonymized text-based metrics (headcount, age, gender, and movement), and purged immediately. Because raw video footage is never recorded or stored, data-breach risks are eliminated at the source—letting operators run the system without privacy impact assessments or regulatory red tape.

클릭하여 확대

Turn Your Existing CCTV Cameras into Smart Spatial Analysts

The 63 Building rooftop deployment proves a critical point: achieving smart, safety-first facility management doesn’t require expensive hardware overhauls. Simply layering a proven AI analytics engine onto the CCTV cameras already watching over your property delivers the most cost-effective and accurate occupancy management system available.

"We’re on a tight budget—can we start small with just one or two channels?"
"Will age and gender analytics actually work with the older CCTV cameras already installed in our building?"
"Can we get accurate foot-traffic counts and automatic age breakdowns without construction downtime or wasted budget?"

If any of these questions sound familiar, reach out to AIPro Co., Ltd. today. We’ll evaluate your existing camera infrastructure, recommend the ideal channel configuration for your site, and give you a clear, practical roadmap to get started.

✔ Automated People Counting Using Existing CCTV

✔ Real-Time Crowd & Occupancy Safety Management

✔ Privacy-Safe (De-Identified) Visitor Analytics System

✔ Building Rooftop Access & Occupancy Control Solution