Content-based image retrieval

Content-based image retrieval (CBIR) aims to find images in a large database that are visually similar to a given query image, relying on the image content itself rather than manual annotations or keywords. In the context of BALaT, this enables researchers and the public to discover artworks that share iconographic, stylistic, or composition similarities. Our approach builds on deep learning feature extractors that map images into a common embedding space, where pairwise similarity can be efficiently measured.