|"Holistic 3D Human Reconstruction In-The-Wild" |
|Konuşmacı ||: || |
Dr. Rıza Alp Güler
|Tarih ||: ||28 Ocak 2020 (Salı) |
|Saat ||: ||13:30 - 14:30 |
|Yer ||: ||Bilgisayar ve Bilişim Fakültesi, |
İdris Yamantürk Konferans Salonu 1303
3D reconstruction from a single RGB image is a fundamentally ill-posed problem, but we perform it routinely when looking at a picture. Prior information about geometry can leverage on multiple cues such as object contours, shading or surface-to-image correspondences, but, maybe the largest contribution to monocular 3D reconstruction comes from semantics: the constrained variability of known object categories can easily resolve the ambiguities in the 3D reconstruction. In this work, we propose to link these separate research threads in a synergistic architecture that combines the powers of the different approaches. We propose a multi-task network comprising 2D, 3D and dense correspondence estimation to drive the 3D reconstruction task. For this, we introduce an iterative refinement method that aligns the model-based 3D estimates of 2D/3D positions with their image-based counterparts delivered by CNNs. We validate our contributions on challenging benchmarks, showing that our method allows us to get both accurate joint and 3D surface estimates while operating at more than 10fps in-the-wild. More information about our approach is available at http://arielai.com/holopose.
Rıza Alp Güler received B.S. and M.S. degrees in Electronics Engineering from Sabanci University, Turkey. He received his PhD degree from École Centrale Paris as part of INRIA and Center for Visual Computing(CVN) group, in 2019. Currently, he is a Post-Doctoral researcher at Imperial College London, Department of Computing. His main research interests are on learning-based models for 3D surface analysis, geometric deep learning, generative and discriminative approaches in human understanding.