In a world where speed and accuracy are the pillars of logistical performance, computer vision provides a concrete response to the challenges of modern warehouses. At XXII, we develop solutions for augmented video analysis that push the limits of perception, particularly through the detection of new objects.

A warehouse is alive: it evolves every day

In a logistical environment, new objects are constantly appearing:

• incomplete or poorly positioned pallets,

• unidentified or damaged parcels,

• equipment left in aisles,

• vehicles moved or poorly parked…

These events, often minor when taken in isolation, can generate delays, picking errors, or even security risks. However, traditional systems based on rigid rules or catalogs of known objects quickly reach their limits. This is where the detection of new objects comes into play.

An AI that sees further: how does it work?

The detection of new objects (or "novelty detection" in English) relies on the continuous analysis of video streams to identify any unusual or unexpected element compared to a reference scene. Unlike simple object recognition (where AI looks for a box or a pallet truck), this technology signals what has never been seen before.

Key ingredients:

• Self-supervised models or with few examples: capable of learning the "normal" without manual labeling.

• Multimodal vision: fusion of video with IoT sensors to contextualize anomalies.

• Real-time digital twin: dynamic visualization of the logistical environment with smart alerts.

What we detect today

Thanks to the advancements of our CORE platform, we have industrialized the automatic detection of several key objects for logistics:

• ✅ Pallets: identification, positioning, detection of anomalies (broken, missing, or moved pallets…).

• 📦 Parcels: real-time tracking, detection of forgotten, misrouted, or damaged parcels.

• 🦺 PPE (Personal Protective Equipment): verification of the wearing of vests, helmets, or safety gloves in risk areas.

This triple detection enhances productivity, traceability, and especially the safety of operations.

Concrete use cases in logistics

At XXII, we have integrated these functions into our solution to provide logistical players with a proactive view of their operations. For example:

• A pallet placed in a traffic area is immediately detected as a risk.

• A parcel that has fallen or has not been scanned triggers a real-time alert.

• An operator without a safety helmet in a regulated area is automatically spotted.

The benefits for smart warehouses

The impact is immediate:

• Reduction in incidents and downtime.

• Fine traceability of goods movements.

• Enhanced safety through continuous monitoring of PPE.

• Productivity gains through automated detection.

And tomorrow?

The arrival of visual language models (VLM), which we are already integrating at XXII, allows us to go further: not only does the AI detect an object, but it can describe it, explain it, and suggest an action.

Example: “A large parcel has been present for 27 minutes in the evacuation area. Intervention recommended.”

This level of understanding paves the way for a growing autonomy of logistical systems, with human supervision refocused on critical cases.