Identification of Art Styles of Tectonic Maps Using Machine Learning

The paper aims to verify the possibilities of the Orange software for defining and identifying artistic styles of tectonic maps using machine learning techniques. A set of tectonic maps obtained from online sources was tested so that the selection of maps was not influenced by the data capture method. The collected maps differ in color and design mainly. The maps evaluated were included in the analysis of the artistic style in the Orange software as pictures. The Painters embedder was applied, which has the most significant potential for clustering maps according to the artistic style of all embedders in the Orange. All maps were cartographically described and subjected to a “map-use” experiment to obtain a subjective evaluation of artistic style by 30 map readers. The obtained evaluation was compared with the assessment of the artistic style using neural networks from Orange, which was used to determine how maps are grouped according to art map style using already integrated neural networks with hierarchical and non-hierarchical clustering methods. The results of the user evaluation in the experiment were compared with the results of the Orange evaluation. The classification of tectonic maps according to the map style and the results from the Orange embedder were compared. The paper results in a recommendation for the creation of a neural network to evaluate the artistic style of not only tectonic maps. Thanks to the created neural network, searching for maps of an art style identifies easily. The paper reveals an original way of identifying the artistic style of the map to support the interpretation of information in the map.