Generate novel views for partially occluded objects using a data-driven approach

• Unité de recherche: INNOSUISSE
• Numéro de projet: 15599.1 PFES-ES
• Titre du projet anglais: Generate novel views for partially occluded objects using a data-driven approach

Description succincte (Allemand)

Generate novel views for partially occluded objects using a data-driven approach

Description succincte (Anglais)

Generate novel views for partially occluded objects using a data-driven approach

Résumé des résultats (Abstract) (Allemand)

In this CTI project, ETH Zurich and LiberoVision AG will investigate and design novel methods to generate high quality free view-point point video with the focus on a data-driven approach. We propose to design a new approach handling occlusions by introducing novel ways for geometry- and texture completion. The data-driven solution will also allow improving the resolution of player textures. This will open new markets in sports analysis.

Résumé des résultats (Abstract) (Anglais)

In this CTI project, ETH Zurich and LiberoVision AG will investigate and design novel methods to generate high quality free view-point point video with the focus on a data-driven approach. We propose to design a new approach handling occlusions by introducing novel ways for geometry- and texture completion. The data-driven solution will also allow improving the resolution of player textures. This will open new markets in sports analysis.