Lawrence Technological University researchers have developed an algorithm that demonstrates computers can be made to understand art in a way very similar to how art historians perform an analysis.
During testing, the researchers used about 1,000 paintings by 34 well-known artists, and programmed the algorithm to analyze the similarity between them based on the visual content of the paintings. The algorithm produced a network of similarities between painters that runs parallel to the analysis of art historians. The analysis showed that the system was able to identify the differences between artistic styles, and designated painters as either classical or modern. The algorithm also was able to identify subgroups within the classical and modern styles.
The experiment was performed by extracting 4,027 numerical image context descriptors, which are numbers that quantitatively mirror the content of the image, from each painting. This enables the computer to reflect many aspects of the visual content, and use pattern recognition and statistical methods to spot complex patterns of similarities and dissimilarities between the artistic styles and then measure the similarities. The researchers say the technology can outperform untrained humans in analyzing fine art.
From Lawrence Technological University
View Full Article
No entries found