In the seminar, we will study methods for depth estimation and 3D reconstruction from images. This is an important problem within computer vision with wide-ranging applications from autonomous driving to human-computer-interaction. As for most other problems in computer vision, state-of-the-art approaches are typically based on deep neural networks ("deep learning"). We will therefore focus on deep learning approaches.
Throughout the seminar, we will study different depth estimation and 3D reconstruction problems, including several approaches for estimating 3D-human pose from images. We will discuss problem settings, deep learning models, training techniques, and empirical results.