Three-dimensional object reconstruction using neural networks and machine learning
Univ.-Prof. Dipl.-Ing. Dr. Lukas Winiwarter
Inhalt:
Neural Radiance Fields (NeRFs) and 3D Gaussian Splatting (3DGS) are on an all-time-high in computer vision. In contrast to conventional multi-view stereo reconstruction, these are based on modeling the distribution of optical density and color in object space. Through machine learning, convergent observations are combined to estimate these distributions. To what extent are these methods also suitable for topographic data, e.g., captured by uncrewed areal vehicles? This presentation will showcase application examples and demonstrate the advantages and disadvantages of these new methods and their suitability for photogrammetry in various applications.
Vortragender:
Lukas Winiwarter has been Professor of Geodesy at the University of Innsbruck since February 2024. His research interests include photogrammetry, geoinformation and the combination with and use of artificial intelligence. He develops and uses innovative methods to record and process three-dimensional data and extract information.