Research on 3D reconstruction has long been focused on recovering 3D information from multi-view images captured in ideal conditions. However, the assumption of ideal acquisition conditions severely limits the deployment possibilities for reconstruction systems, as typically several external factors need to be controlled, intrusive capturing devices have to be used or complex hardware setups need to be operated to acquire image data suitable for 3D reconstruction. In contrast, 3D reconstruction in unconstrained settings (referred to as 3D reconstruction in the wild) usually imposes only little restrictions on the data acquisition procedure and/or on data capturing environments, but, therefore, represents a far more challenging task.
The goal of this workshop is to foster the development of 3D reconstruction techniques that are robust and realtime, and consequently perform well on a variety of environments with different characteristics. Toward this goal, we are interested in all parts of 3D reconstruction techniques ranging from multi-camera calibration, feature extraction, matching, data fusion, depth learning, and meshing techniques to 3D modeling approaches capable of operating on image data captured in the wild. Topics of interest include, but are not limited to: