Information Discovery and Images A Case Study of Google Photos

Images become more important as carriers of information. But information retrieval and discovery is still mainly based on words and text. In recent years progress is made in exploiting image-based information also. Searching for images using a text query is now already a classical method. Search by image or reverse image search is a more recent method in which the query consists not of text but of an image. This approach allows us to find and reveal exact and modified copies of a known image that is used as source image in the query. Since a few years, this method even makes progress in order to find images on the World Wide Web (WWW), which are not only visually similar, but even semantically related to the query/source image; the same search action can also reveal related text information. All this has been demonstrated in previous tests by this author. A leading developer in this area is the company Google. More recently Google has changed their service for storage of photos by creating the new service named Google Photos; there also, the system applies the improving methods of automatic semantic analysis to the submitted photos. This results in automatic classification/categorization/annotation/tagging of photos, according to their contents. Here a case study of this feature is reported. It turns out that categories are created with high precision, as hoped; however, specificity is only low, as expected. This demonstrates at least that automatic semantic analysis of images is growing in importance. So managers of a digital library that includes images should keep an eye on this evolution in order to maximize the information discovery process of their users.