Suggestion, alternatives and quantity
The result remains under the user’s control.
PackQR 1.3 recognition guide
A helpful shortcut—not perfect recognition.
AI Pack Scan can suggest household-item names and quantities from a photograph. It is designed to reduce repetitive typing during a move, not to identify every object correctly.
✓ On-device Apple Vision✓ Works without internet✓ No hosted large-AI subscription✓ Every result editable
The honest expectation
Useful assistance with clear limits.
PackQR has been developed and tested with a varied collection of household-item examples. That helps the built-in recognition cover many ordinary packing situations, but it is not a guaranteed catalogue.
The same kind of item can look very different between homes, brands and photographs. A mug may be recognised as a mug, cup, plate or something broader. An unusual object may not receive a useful suggestion at all. The user can always select an alternative, type a custom name and change the quantity.
Please review before saving.
A wrong or uncertain suggestion is an expected limitation of the feature. PackQR deliberately keeps the review step visible instead of pretending that recognition is infallible.
Saved into the normal inventory
The reviewed item and photo remain local.
Why local matters
Recognition without sending the move to a large cloud model.
Household photographs and inventories can reveal a great deal about a home. PackQR keeps the core recognition workflow on the iPhone or iPad using Apple Vision.
An active internet connection is not required. Item photographs are not uploaded to a PackQR server or hosted large-AI recognition service. The feature does not depend on an ongoing subscription to an external AI-model API.
No PackQR recognition cloudThe photo is analysed on the device.
No monthly AI-model dependencyThe pay-once app does not need an external model subscription to keep recognising.
Useful during a stressful moveCore scanning and recognition remain available in low-signal rooms, storage units and moving vans.
Your review remains finalPackQR saves the name and quantity that you approve.
Why results vary
Photography and the environment matter.
Recognition sees the photograph rather than the object as a person understands it. Small changes can alter the available visual evidence.
Ambient lightingVery dark rooms, hard backlighting, colour casts and uneven light can hide shape and surface detail.
Glare and reflectionsGlass, polished metal and glossy packaging may reflect windows or lights instead of showing the object clearly.
Camera angleAn unusual view can make familiar items resemble another category.
Distance and framingObjects that are small in the frame provide less useful detail.
Background and obstructionBusy surfaces, overlapping items and hands covering key features can reduce confidence.
Mixed groupsSeveral unrelated objects are harder to represent as one saved inventory row and quantity.
For the best chance of a useful suggestion
Photograph one item or one related group at a time, keep it large in the frame, use even light and try another angle when the first result is uncertain.
Optional personalised learning
Similar suggestions may improve as packing continues.
When learning is enabled, PackQR can create a compact visual representation after the user deliberately corrects or teaches a saved item. It does not silently learn from every photograph.
Later photographs can be compared with those locally stored examples. Several useful views of a similar object can make the learned suggestion stronger. However, one example is not a promise, and even a learned label may remain an alternative when the new photograph is different or conflicts with other evidence.
- Learning examples stay on the device.
- They remain linked to the source item and photograph.
- Users can pause learning while continuing to use existing suggestions.
- Users can stop using learned suggestions without deleting the existing library.
- Linked learning is removed when its source item or photograph is deleted.
- Learning can improve similar cases, but it will not make recognition perfect.
Recognition questions
What users should know.
Is AI Pack Scan expected to recognise everything?
No. It is a practical shortcut, not a universal or guaranteed identification system.
Can I correct a result?
Yes. Choose another suggestion, type a custom name and change the quantity before saving.
Does it need an internet connection?
No. The core recognition workflow runs on the device after the app is installed.
Are my item photographs sent to an external AI service?
No. PackQR uses Apple Vision locally and does not upload item photographs to a hosted large-AI recognition model.
Why was an obvious object recognised incorrectly?
The model only receives visual evidence from the photograph. Lighting, angle, reflection, distance, background, obstruction and object variation can all influence the result.
Will learning fix every future result?
No. Deliberately taught examples may make similar suggestions more useful over time, but learned recognition remains an aid rather than a guarantee.
Keep packing
Use the shortcut when it helps. Type the answer when it does not.
PackQR is designed so uncertain recognition never blocks the moving inventory workflow.