Hasan Onur ÇalışkanVisual product recognition
Mobile application work aimed at helping store staff find shoe products from images.
The business need
Finding a shoe shown by a customer in a catalog can be slow, particularly when models look similar. Camera angles, lighting and product variants make matching harder.
My area of work
My shoe recognition mobile application work combines computer vision with retail use. The focus is supporting staff in product search and verification, not making uncontrolled stock or sales decisions.
Proposed pilot workflow
This workflow illustrates a technical approach. It is not a recording of a working product or a completed customer implementation.
Build a pilot catalog with authorized images and product identifiers, clearly defining which products it includes.
Use photos from different angles and lighting conditions. Real customer data is not needed for the first pilot.
Present product candidates for review rather than one supposedly certain answer. Account for products outside the catalog.
A staff member confirms the model and variant. Inventory queries are scoped as a separate integration.
How should the pilot be evaluated?
- Measure how often the correct product appears among the first suggestions.
- Record false matches for products outside the catalog.
- Test lighting, angle and similar-model robustness using new photos.
Limits and human review
- A photo is not always enough to identify an exact model, size or color variant.
- This page makes no claim of verified accuracy or live inventory integration.
Evidence and current status
No actual product recording or verified performance measurement is published on this page. Business-specific results can only be shared after an authorized pilot and measurement.