From Football Player Tracking to Shelf Intelligence: How Vision AI Decides the Moments that Matter

Warning: This blog contains many football-related puns. See if you can spot them all. For our American readers: this blog is about football, the one played with feet. We acknowledge the debate is ongoing...

When you are watching a football match this summer, AI will be watching too.

But it’s not watching with the same emotional attachment, ups and downs (and eventual disappointment) as we are.

It’s converting the beautiful game into structured data in the same way retailers are turning their shelf setups into sales.

New player tracking vision AI technology

Earlier this year, football’s governing body announced it would deploy vision AI technology to improve automated offside decision-making.

The setup is designed to give poor decisions the red card.

Every one of the 1,248 players at the tournament stepped into a scanning booth lined with fiducial markers (reference points that tell the cameras exactly where they are in the room) and was scanned to create a full 3D avatar.

Image courtesy of Lenovo

The result is a body model specific to that player, with accurate dimensions for every limb.

Sixteen cameras in the stadium will capture their position 50 times per second. Twenty-nine points on their bodies are being tracked simultaneously.

A sensor inside the ball fires data 500 times per second to pinpoint the exact moment it leaves their teammate's boot, so their body position can be checked against the offside line.

Decisions will be made within seconds.

Shelf intelligence vision AI technology

Now step from the stadium to the grocery store. Vision AI technology used to secure offside decisions is being applied to secure the perfect hat-trick of in-stock, and correctly placed and priced inventory.

Generic shapes are not good enough for football, which is why every player has been scanned. Shelf intelligence reached the same conclusion years ago. A generic bounding box around a product-facing view is sufficient for detecting whether something large is present.

It is not fine for detecting whether a specific SKU is correctly placed, the right way round, and not substituted with the wrong product. You need a trained model of the actual object to avoid a penalty.

The cameras might be slightly different, mobile cameras and autonomous robots have joined the team, but the tracking logic remains the same.

The system will compare on-shelf availability levels with what should be there. The moment an error is spotted, a prioritized alert fires.

A store associate gets an assist on the sale by placing a product right back on the shelf where it belongs before the customer notices.

Two different arenas. Same idea. Because in both cases:

• Cameras observe reality
• AI identifies objects
• Objects become data points
• A digital twin is created
• Rules-based decisions are generated

Comparing the two

Here's how the vision AI tech stacks compare.

Technology dimension	Football Match Monitoring	Shelf Intelligence
Camera setup	Fixed cameras mounted under the stadium roof, covering the full pitch	Different fixed-position cameras, mobile and/or robots, covering the full store
Object detection	Detects and tracks each player as a unique object in the scene.	Detects and tracks each SKU as a unique object on the shelf
Keypoint/model tracking	29 skeletal keypoints per player, including limbs, joints, and extremities	SKU-level product models trained on design, dimensions, and orientation.
Reference comparison	Compares player position against the offside line - where they are allowed to be	Compares shelf state against the formation - what should be there and where
State change detection	Identifies the exact millisecond the ball leaves a player's foot	Identifies when a product is out of stock, misplaced, or priced incorrectly.
Automated alerting	Offside alert sent to VAR officials the moment a violation is detected	Out-of-stock or compliance prioritized alerts sent to store staff for action.
Decisive action	The officials have clarity on whether it’s offside or not, and the correct action is taken.	The store associates are guided on what needs to be fixed and complete the task quickly.

Football’s new system works because the geometry is fixed. The pitch stays the same, and the cameras capture the same image. Every position the system measures is referenced to a stable frame. That is what makes centimeter-accurate offside calls possible.

Shelf intelligence works for exactly the same reason. The cameras take the same picture. The shelf doesn’t move. The planogram defines what correct looks like. Everything vision AI sees is seen in context.

Formation deviations from the reference state are a simple matter of maths.

The key difference

There is one honest difference worth calling out. Football's system is built to catch a single moment. The ball leaves the boot, and the call is made. Speed is the whole game.

Shelf intelligence is built for something harder. The shelf degrades continuously throughout the entire match day. Products sell. Errors creep in. Prices drift. No single millisecond defines the problem. The cost builds quietly over hours.

That changes what the AI needs to do. Both systems have had to solve the same prioritization challenge, but from different angles. An official doesn't need to know every player's position every second. They need to know when something crosses a line.

A store associate doesn't need a notification every time a shelf image is captured or when something is out of place. They need to know when a best-selling product has been left on the subs bench, or when a high-velocity SKU drops below threshold at peak hour.

The goal in both cases is the same: surface the right call at the right moment, without flooding the officials with noise. A referee buried in alerts makes bad decisions. So does a store associate.

The beautiful game. The perfect shelf.

This summer’s tournament demonstrates that fixed cameras, trained object models, and a reference state to compare against are enough to make decisions that were previously impossible at the required speed and accuracy.

Retailers have been sitting on the same insight for years. The shelf is a pitch. The planogram is the offside line. The product is the player. The question the camera is always asking is the same one.

Is it where it should be?

For shoppers and football teams alike, there is only one acceptable result. Taking home exactly what you came for.

Cameras, object detection, planogram comparison, and automated alerting for the moments that matter.

ShelfView Shelf Intelligence

Learn more