Search by what you see
A visual-search service that helps visitors find similar furniture from a photograph.
A project from peoly’s published portfolio.
- 01Photo query
- 02Visual features
- 03Similar products
The challenge: describe a product without words
A furniture aggregation platform wanted to extend its search beyond text. Visitors needed to find similar products by uploading a photograph they had taken or saved.
From an image to similar products
peoly designed and developed the service with a team combining machine-learning specialists and senior software engineers. The architecture brought together object detection, extraction of visual feature vectors and approximate nearest-neighbour search.
The detector identifies the relevant object in an image. Its visual representation provides a basis for comparison, allowing the search service to return similar catalogue items rather than relying only on a written description.
A service within the existing platform
The solution was delivered as a separate microservice for the customer’s Kubernetes infrastructure. The work included the communication model and integration with services already used by the platform.
The case demonstrates how an AI capability becomes part of an established product: the model and search pipeline are connected to the application that customers already use. For another catalogue, the product categories, image material and meaning of a useful match define the project.
A similar challenge in your operation?
Tell us about your data, devices and the workflow you want to improve.