All case studies
  • Artificial intelligence
  • Digital healthcare

Asset inventory with image recognition

An app to record and verify assets in the field: from a photo it reads the labels, suggests the asset type and flags similar items already recorded. All on the phone, even without a network.

The starting point

In a large organisation, such as a hospital, there are thousands of assets to inventory: beds, wheelchairs, equipment, furniture. They sit in different buildings, floors and rooms, and many carry several labels, from the organisation, clinical engineering or third-party owners, or none at all.

Whoever carries out the inventory must recognise each item, find it in the existing list or record it as new, read codes that are often small or worn, and place it in the right location. Every mistake ends up in the file to be delivered.

How it works in the field

  1. 1Where am IThe operator picks the location from guided lists: site, building, floor, department, room.
  2. 2A photoThey photograph the asset and its labels: inventory, clinical engineering, serial number, third-party owners.
  3. 3AI suggestsCodes read, asset type and similar items already recorded: the operator confirms or corrects.
  4. 4Label and goIf needed they print the new label on the portable printer; data syncs as soon as there is a network.

What the artificial intelligence does

Three models work on the same photo, each with a specific task. They suggest, they do not decide: the operator has the last word.

Reads the labelsText recognition extracts inventory codes, serial numbers and registration numbers, even from small or damaged labels.
Recognises the assetAn image classifier suggests the category, for example bed, wheelchair or infusion pump.
Finds similar itemsIt compares the photo with those already collected and shows the most similar assets: it helps avoid duplicates and find items that have moved.

Lightweight models, trained on the context

The classifier starts from a compact neural network, already trained on a large image archive and designed for mobile devices. We specialise it with photos of the client’s assets, organised by context: each organisation can have its own categories.

The trained model is converted into a standard format and runs directly inside the app. When photos and categories are added, it is retrained and redistributed to the devices.

The system works out the rest

The operator focuses on the asset and where it is. The fields that the organisation’s file derives from formulas and rules are calculated by the platform.

  1. Initial listThe portal imports from Excel the assets already recorded and the lists of valid locations and cost centres.
  2. Guided surveyNo free text where a code is needed: only allowed values, with lists that narrow level by level.
  3. Automatic derived fieldsManagement type, acquisition, reason code and funding source are derived from the codes collected, according to the agreed rules.
  4. Final fileAt the end of the campaign the portal generates the complete, compliant file, ready for the organisation’s asset system.

What changes

Less time per asset, fewer codes copied by hand and fewer duplicates. The operator confirms what the app suggests instead of typing everything from scratch, coordinators see progress department by department, and at the end the file to deliver is already complete, with photos backing every record.

More case studies

All case studies

We are always curious to hear how your needs are changing

Half an hour on Teams: you tell us where you are heading, we tell you frankly whether and how we can be of use.

Book a meeting