June 17, 2022
The IEEE/CVF Computer Vision and Pattern Recognition (CVPR 2022) has accepted a paper by researchers from the "Vision and Imaging Technologies" department of the Fraunhofer Heinrich Hertz Institute (HHI). The paper by Markus Worchel, Rodrigo Diaz, Weiwen Hu, Oliver Schreer, Ingo Feldmann and Peter Eisert was written within the framework of the EU project INVICTUS . It covers the highly accurate and performant 3D reconstruction of objects.
In the digital age, avatars have become ubiquitous. Given the huge market for avatar-based digital entertainment, the EU-funded project INVICTUS (Innovative Volumetric Capture and Editing Tools for Ubiquitous Storytelling) is developing digital design tools to create animatable volumetric videos. These tools enable the creation, editing and re-animation of dynamic 3D representations of virtual humans in high quality. In this context, Fraunhofer HHI focuses on the development of new and efficient technologies for volumetric capture, post-processing tools for volumetric videos and the realistic synthesis of new 3D sequences via neural rendering.
The research paper "Multi-View Mesh Reconstruction with Neural Deferred Shading" by employees of Fraunhofer HHI, which was written in this context, deals with the generation of high-quality 3D models of objects and persons from images. By using differential renderers as well as AI-assisted analysis of reflection properties of surfaces, it is possible to reconstruct finest details and modify material properties of virtual objects. The methods presented in the paper find application in the improved generation of volumetric videos and in the generation of 3D assets for virtual environments.
The paper will be presented at this year's IEEE/CVF CVPR, which will be held in June in New Orleans, USA. The IEEE/CVF CVPR is the premier annual event on computer vision and includes the main conference as well as several concurrent workshops and courses.