Development of a color-based search engine for the database of the Hungarian National Museum

Partner:

Hungarian National Museum’s National Centre of Museological Methodology and Information​

Domain

Custom solution

Technologies

Color clustering algorithm

The Hungarian National Museum’s National Centre of Museological Methodology and Information​ came to Régens looking for an artificial intelligence-based solution that enables color-based searches in the museums' aggregated database.

Challange

The MuseuMap portal collects the public elements of the digital database of museum institutions registered to the program and makes them available on a single interface. The Hungarian National Museum’s National Centre of Museological Methodology and Information wanted to facilitate the search for these contents with the most modern technologies, so it charged Régens with the development of an artificial intelligence-based search engine that analyzes the elements of the database and categorizes them based on colors, implementing the possibility of color-based filtering of works of art.

Solution

Régens has developed a unique color clustering algorithm to implement the desired color-based filtering. This algorithm detects the colors found on the images, and then the system's artificial intelligence algorithms decide which of the mapped colors are the main colors to be displayed and with what priority they appear on the image.

Result

The developed solution is available to users under the filter conditions of the MuseuMap result list and also in a separate menu item. After selecting a specific color, the system lists the works of art for which it detected that color as a main color. A predefined color palette with a total of 81 shades is available for selection. The system can sort the results by the percentage of this color on the images. In addition, after filtering for a given color, the system also allows us to specify additional colors and further filter the results based on all selected colors. In the filter criteria, it offers the colors that are most common in search results, and then sorts the results by examining the two (or more) colors together. Due to the large number and variety of images, color filtering greatly facilitates browsing and targeted searching.