U ponedjeljak, 28.3.2022. s početkom u 11h, u Sivoj vijećnici na Fakultetu elektrotehnike i računarstva, Unska 3, Zagreb, prof. dr. rer. nat. Stefan Gumhold, sa TU Dresden, Njemačka održat će predavanje na temu Model Based Human Shape and Pose.
U sklopu svojih aktivnosti predavanje zajednički organiziraju Znanstveni centar izvrsnosti za znanost o podatcima i kooperativne sustave (ZCI-DATACROSS), Laboratorij za naprednu 3D rekonstrukciju i registraciju površina (SHARK lab) Fakulteta elektrotehnike računarstva, Odjel za obradu signala Hrvatske sekcije IEEE, Hrvatsko društvo za biomedicinsko inženjerstvo i medicinsku fiziku te Odjel sustava i kibernetike HATZ-a.
Short Bio: Stefan Gumhold is a full professor in Computer Graphics and Visualization at TU Dresden in Germany. His research areas are 3D scanning and geometry processing, real-time rendering and scientific visualization, as well as scene understanding. He received his PhD from the University of Tuebingen on the topic of Mesh Compression and visited the Technion in Haifa, Israel, the Sci-Group in Salt-Lake City, USA and the University of Granada, Spain as a postdoc researcher. He is PI in several large project like the cluster of excellence “Centre for Tactile Internet with Human-in-the-Loop”, “Center for Scalable Data Analytics and Artificial Intelligence”, “6G-life cluster” and the “Center for Perspicious Computing”. Furthermore, he chairs the environmental commission that advises the Rectorate on the ecological sustainable development of the TU Dresden.
Model Based Human Shape and Pose
The talk starts with a short overview of the research topics currently investigated at the chair of Computer Graphics and Visualization including immersive labeling, GPU ray-casting for scientific visualization and medical image fusion. The main part of the talk will give an introduction to parametric shape and pose models for humans that will be used in the 6G-life cluster to support telepresence in immersive collaboration. After a short overview of the definition and construction of parametric models like SMPL-X, recent pipelines that allow fitting of the 3D human model to photographs, videos and rgbd-scans are covered with a focus on real-time capability and avatar creation.