The main idea of the thesis is to design and implement a tool that identifies and retrieves sneakers based on a photograph. The use of machine learning models or APIs built on top of such models is expected.

The student will first research and evaluate the current state of the art in the field of visual search and 3D object identification based on visual characteristics. Furthermore, the student will analyze and identify which visual parameters are most important for effective sneaker identification.

Then, the student will create a web application into which sneaker catalogs (based on existing datasets) can be uploaded. Users will be able to create their own sneaker collections, which will be matched to items in the catalog. As a result, relevant metadata will be automatically assigned to each product.

The search and matching process will be primarily driven by visual similarity based on a photo of a sneaker. The catalog items as well as the searched object may each be represented by multiple images from various angles or zoom level.