Search

This article discusses a number of modern techniques used for the analysis of works of art. The most widely used approaches as well as lesser known ones are outlined in terms of their applications and the kind of information on the condition of artworks that can be extracted. Special attention is paid to the method of thermographic analysis of works of pictorial art. The principles of the technique, various computational approaches, and safety concerns are discussed. A set of examples is provided for the demonstration of the capabilities of thermographic assessment, including a range of real canvas and panel paintings exhibited in museums and in private collections. PACS Nos.: 87.63.Hg, 78.20.nb, 81.70.Pg, 07.20.-n. Nous discutons ici de techniques modernes utilisees pour analyser les ceuvres d'art. Nous presentons les methodes les plus connues comme les moins connues, selon leurs applications et la sorte d'information qu'on peut extraire sur les conditions de l'ceuvre d'art. Nous portons une attention speciale a l'analyse thermographique d'ceuvres picturales. Nous discutons les principes de la technique, des differentes approches informatiques et des soucis de securite. Nous presentons un ensemble d'exemples pour illustrer les capacites de l'evaluation thermographique, incluant plusieurs veritables tableaux et panneaux peints exposes dans des musees et des collections privees. [Traduit par la Redaction], 1. Introduction In the last several decades there has been increased interest in the preservation and conservation of objects of cultural heritage. The number of publications presented at international conferences [...]

This paper covers the main scientific results and achievements of the Research Center for Applied Artificial Intelligence Systems of the Moscow Institute of Physics and Technology. Accomplishments in key research areas, such as (i) "Natural language analysis based on artificial intelligence methods" and (ii) "Artificial intelligence for robotics and uncrewed vehicle control," are described. Specifically, within (i), multimodal and recommender models have been studied and text has been shown to be promising from the point of view of modality fusion: the prevailing majority of successful multimodal products work with text modality, and embeddings of any modality can often be reduced to a text embedding. At the same time, applications of multimodal models are obviously little developed: the ability to formulate and solve particular business tasks with their help is in its initial state. Within area (ii), techniques and algorithms are developed for designing a control subsystem for a robotic vehicle intended for visual simultaneous localization and mapping in real time, which improve the quality of navigation under various weather conditions in various environments (urban, rural, highway, etc.). The implementation of this project would simplify the initial prototyping of systems for navigation, computer vision, and positioning of unmanned robotic complexes and devices due to rapid data processing. Additionally, the development of bipedal anthropomorphic robots is underway: research in this scientific-engineering area is actively conducted around the world, numerous scientific works are published, and various competitions are held. To achieve the multifunctionality and flexibility required for operation in a human-friendly environment, a robot should have a design and mechanics maximally similar to parameters human. These requirements are most closely met in bipedal anthropomorphic robots. The concept of a robot design satisfying these requirements has been developed, and its detailed design for prototyping has begun. Additionally, key publications on the presented topics are described, and educational activities of the Center are surveyed. [ABSTRACT FROM AUTHOR]