Can AI cameras detect racking damage?

The one result that's actually been verified

In 2022, researchers at the University of Huddersfield published a peer-reviewed study in the journal Sensors, describing a YOLOv7-based computer-vision framework for automated, edge-device pallet-racking defect inspection. The framework achieved a mean average precision of 91.1% in their testing. A related follow-on paper from the same research group, applying a YOLOv5-variant selection method to the same problem, was published in 2023.

That's the strongest evidence found anywhere in the research behind this page that computer vision can genuinely detect racking damage. It's worth being precise about what it is and isn't: a peer-reviewed research result, published with a stated methodology and accuracy figure that other researchers can scrutinise. It is not a shipped commercial product, and it doesn't tell you what any given vendor's camera actually achieves in a working warehouse.

What the commercial vendors are actually claiming

Several commercial vendors sell AI camera products marketed at warehouse safety and racking condition, including viAct, OxMaint and iFactory, covering claims like detecting overloaded racks, unsafe stacking height, unauthorised access to high racking, and forklift impact damage. One vendor's own site states figures like "up to 65% reduction in inventory damage" and "up to 75% reduction in forklift near-misses" from what it describes as real deployments.

Every one of those figures is self-published on the vendor's own site. None names a customer, states a methodology, or points to independent audit or verification. That's not a technicality — it's the difference between a claim and an established fact, and it should be treated as exactly that: a vendor's claim about its own product, not a verified result.

Why that gap matters for a real buying decision

A single peer-reviewed study with a strong accuracy figure is genuinely promising evidence that the underlying technique works. It is not evidence that any specific commercial product, deployed in any specific Australian warehouse, will reliably catch the kind of damage that actually causes a racking failure. Anyone evaluating a vendor's AI-camera product for racking condition should ask directly for independently audited results, not accept a vendor's own marketing figures as proof, in the same way you'd ask a vendor to substantiate any other safety-relevant performance claim.

It's worth separating this narrow claim from a broader, better-evidenced one: analyst firm Gartner has forecast wider adoption of AI-enabled vision systems in warehousing generally, aimed largely at replacing manual barcode scanning for cycle counting and general safety monitoring. That forecast is about counting stock and watching for hazards, not about detecting structural racking damage specifically — conflating the two is exactly the kind of overreach this page is trying to avoid. See where AI is actually earning its keep across the wider warehouse for that broader picture, not just racking cameras.

What this technology cannot replace

No camera system, proven or not, replaces the judgement of a competent person carrying out a formal racking inspection. Even a genuinely capable computer-vision system is, at best, a way to flag visible surface damage for a human to assess — it isn't equipped to make the structural-engineering judgement about load capacity, deflection or safe continued use that the person carrying out a competent-person inspection under AS 4084 is actually for. Nothing in this research supports treating a camera, however accurate, as a substitute for a proper inspection, and this page isn't suggesting one.

The honest summary

One independently verifiable result exists for AI detecting pallet-racking damage specifically: a 2022 peer-reviewed academic study. Everything else currently marketed in this category, at the time this research was done, is vendor marketing with no independent verification behind it. That's not a reason to dismiss the category outright — the underlying research is real and promising — but it is a reason to be sceptical of any specific product claim until it comes with evidence a customer, not just the vendor, can check.

Common questions

Has AI actually been proven to detect racking damage?

In a research setting, yes — a 2022 peer-reviewed study from the University of Huddersfield demonstrated a YOLOv7-based computer-vision framework detecting pallet-racking defects with a stated 91.1% mean average precision. That's a research result, not a verified claim about any commercial product on the market.

Can I trust vendor claims about AI racking-damage cameras?

Treat them as marketing claims until proven otherwise. The vendor performance figures found in this research were self-published, with no named customer, no stated methodology and no independent audit — ask any vendor directly for independently verified evidence before relying on a product's stated performance.

Can an AI camera replace a racking inspection?

No. Even a genuinely accurate computer-vision system flags visible damage; it doesn't carry out the structural-engineering judgement a competent-person inspection under AS 4084 is for. Nothing in this research supports using one as a substitute for the other.

Is computer-vision racking inspection likely to improve?

The underlying academic research is genuinely promising, and it's a plausible near-future category. But at the time this research was carried out, the gap between one proven academic result and multiple unverified commercial claims was real, and worth checking again before relying on any specific product's marketing.

Read next

• how an inspection is actually run
• a report you can act on
• where AI is genuinely deployed in a warehouse
• machine learning for demand forecasting and slotting
• rack-supported clad-rack structure
• floor flatness for automated equipment