Echtzeitdatenpipeline für Ki-Verbessernde Rahmen

Details

Veröffentlichungs-Nr.

EP4811796

Anmeldetag

18. März 2025

Veröffentlichung

23. September 2026

Rechtsraum

EP

Abstract

Example embodiments relate to a computer-implemented method for generating a real-time stream of AI-enhanced video frames on a graphics processing unit, GPU, cluster (211); wherein the GPU cluster comprises a processing unit (212) and a plurality of GPUs (220, 230, 240); and wherein the respective GPUs are configured to support one or more GPU-accelerated containers (221, 222, 231, 232, 233, 241) each running a generative AI-model asynchronously to enhance a video frame into an AI-enhanced video frame; the computer-implemented method comprising, by the processing unit (212), while receiving (218) a live video feed: assigning (202, 214, 215, 216) the respective video frames of the live video feed to available GPU-accelerated containers; outputting (204, 219) a next AI-enhanced video frame to the real-time stream at a specified frame rate, if (203) the next AI-enhanced video frame is available within a threshold interval from outputting the previous AI-enhanced video frame to the real-time stream; else, reducing (205) a performance score of the GPU-accelerated container tasked with enhancing a video frame into the next AI-enhanced video frame; and disabling (206) GPU-accelerated containers with a performance score lower than a minimum performance score defined by a minimum frame rate for the real-time stream.

Anmelder

Firma

Artesis Plantijn Hogeschool Antwerpen

Land

🇧🇪 Belgien