RUS  ENG
Full version
JOURNALS // Doklady Rossijskoj Akademii Nauk. Mathematika, Informatika, Processy Upravlenia // Archive

Dokl. RAN. Math. Inf. Proc. Upr., 2022 Volume 508, Pages 109–110 (Mi danma347)

ADVANCED STUDIES IN ARTIFICIAL INTELLIGENCE AND MACHINE LEARNING

Computer vision approaches in high-fidelity multimedia content synthesis

A. V. Kuznetsov, D. V. Dimitrov, A. Yu. Groshev, P. P. Paramonov, A. A. Mal’tseva

Sber AI, Moscow

Abstract: The development of deep learning technologies inevitably generates new tasks and their solutions in areas, such as computer vision, VR/AR technologies, video analytics, multimodal learning, etc. With increasing availability of high-performance computers, many modern methods and tools for digital data processing become widely applicable, including in personal application research. This tendency can easily be traced in the increasing number of open-source solutions that can easily be executed at well-known resources, such as Google Colab. In this paper, we describe results regarding the development and study of breakthrough technologies of high-quality multimedia synthesis, which have wide applications in tasks, such as face swapping.

Keywords: face swap, GHOST, one shot, image synthesis, video synthesis.

UDC: 004.8

Presented: A. L. Semenov
Received: 28.10.2022
Revised: 28.10.2022
Accepted: 01.11.2022

DOI: 10.31857/S2686954322070141


 English version:
Doklady Mathematics, 2022, 106:suppl. 1, S99–S100

Bibliographic databases:


© Steklov Math. Inst. of RAS, 2024