Artificial intelligence (AI) has revolutionised many fields, including art. Thanks to AI, we can (re)discover similar works of art and deepen our understanding of them and their creation.
AI algorithms, and more specifically artificial neural networks, have the ability to extract and analyse the visual characteristics of works of art, such as colours, shapes, patterns, textures and so on. These neural networks are learned automatically from a large and vast database of works of art or other material. Learning from a vast database enables the network to identify subtle similarities that often escape the human eye. The features that these networks have extracted are used to identify similar works, regardless of their period or artistic movement. These characteristics are in fact in vector form, meaning that each image in the database is associated with a vector of characteristics. The vectors are then compared using distances, and the smallest distances give us the images that are most similar to our query image.
When the network is trained to learn a particular task, such as recognising artistic movements, artistic genres, the date of production or the artist, this adds a filter to the similarity search and facilitates comparative analyses. In effect, the network will focus more on the class to which the image belongs for a given task and will suggest similar works according to the predefined task, since the features extracted will contain important information for recognising the task in question.
The analysis of works of art using AI is not limited to the search for similarities. It also includes understanding the works, their composition, etc.
There are many AI algorithms that can break down a work of art into its constructive elements. For example, an object detector can identify recurring motifs or objects/symbols in works of art and therefore match works with common co-occurrences, but also interpret the work.
The use of AI in the search for similar works of art and the analysis of works enables a new interaction with art. Thanks to its visual analysis and machine-learning capabilities, AI enables us to rediscover and better understand works of art.