CactusXR Uses Surveillance to Estimate Shelf Attention
Auki's retail space platform Cactus demonstrates: without adding new hardware, it uses existing in-store surveillance cameras and generates a digital twin via mobile phones to estimate which shelf locations attract more customer attention.
The system connects surveillance footage to a 3D map of the store, using computer vision to detect pedestrians, estimate body posture and head orientation, and project this onto shelf and product coordinates to create a gaze heatmap. Mobile phones scan shelves and barcodes to complete the twin and product registration, with Auki network QR codes set up in-store as spatial anchor points. The official statement claims it can be self-deployed, with data stored on the merchant's own cloud.
The product page also lists 3D shelf analysis, integration with sales inventory systems, out-of-stock alerts, navigation, and route optimization. External figures include an average daily saving of over 30 minutes for employees, a reduction in walking distance by about 30%, and an increase in shopping basket add-on rates by about 6%. The pilot at Stora COOP Visby on Gotland Island claims approximately 5,000 square meters and over 24,000 SKUs, with handover and search saving each person up to about 45 minutes per day; navigation for self-pickup can reduce walking by about 43%.
The gaze results are termed "estimated attention," which can be aligned with sales data to verify whether displays convert to sales. Merchants can still use the same map to assign tasks with coordinates to employees. Originally intended for security, surveillance is now used as a spatial sensor; the company also provides multi-source access via handheld devices, glasses, and robots.
This serves as a software replacement for dedicated shelf cameras. Buyers are chain stores with existing surveillance that do not wish to install hundreds of shelf cameras, while sellers charge per camera position for shelf visual solutions. Funding shifts from purchasing new cameras to subscriptions and cloud hosting; beneficiaries are operational teams that can use heatmaps to adjust end caps and main aisle displays, while suppliers of dedicated shelf cameras and wiring face pressure. Gaze estimation relies on head orientation rather than eye tracking, with accuracy varying with lens angle and obstructions.
Source: Public Information
ABAB AI Insight
Cactus transforms security surveillance from recording devices into spatial sensors, with the key not being new lenses, but the shelf coordinates generated by mobile scanning. Without twins, cameras only know when someone walks by; with SKU coordinates, they can translate "head facing that way" into "this end cap was looked at." Auki uses QR codes to lay anchor points, reducing the cost of laser radar entry into stores, with 24-hour self-deployment aimed at store managers rather than integrators. Visby Digital sells search and handover, while the gaze function is a new layer added to the same map.
The capital path is to re-monetize already sunk surveillance assets. Chain store cameras are already installed, with incremental value coming from software and cloud. Self-hosted cloud addresses retail concerns about customer flow video outflow. Sales integration allows gaze heatmaps to align with SKU gross margins, turning display experiments from anecdotal to verifiable placements. Handheld scanning replaces dedicated shelf cameras, avoiding capital expenditures of "one machine per shelf position."
It benchmarks against visual checkout solutions like Standard Cognition and Amazon Go, as well as traditional shelf camera inventory checks. Cactus is in an expansion phase using digital twins to leverage existing security, not yet replacing POS. The industry is shifting from "installing new sensors" to "adding coordinate systems to existing sensors."
The structural change is a reconstruction of the industry chain. Security, display, and inventory systems are interconnected through a single 3D map. The mechanism is: whoever controls shelf-level coordinates can transform surveillance video from evidence into product selection input. AI without coordinates can only count foot traffic, while AI with coordinates can optimize layouts.
ABAB News · Law of Cognition
Cameras have been in place; what is lacking are shelf coordinates.
Head orientation is not gaze, but is sufficient for display.
Not installing new hardware allows stores to change layouts the next day.