Daily CVPR - Thursday

The Large-Scale Semantic 3D Reconstruction project was presented both at the orals and as a poster. What it does is automatic generation of semantically annotated 3D city models , starting from large image collections. The idea of semantic 3D reconstruction is to recover the geometry of an observed scene while at the same time also interpreting the scene in terms of semantic object classes (such as buildings, vegetation etc.). While jointly inferring 3D shapes and semantic classes delivers appealing results, up-to-date methods are memory-hungry and computationally expensive. This approach demonstrates how to scale semantic 3D reconstruction up to be able to reconstruct large geographic areas such as cities. The resulting superior 3D models allow for advanced reasoning tasks such as urban planning, navigation, physics-based simulations and the like. This method is not only limited to 3D city modelling: possible further applications include a number of multi-class segmentation problems, e.g. in the field of medical imaging. This six-teamer work is available in full here . CVPR Daily: Thursday 8 Presentations Maros Blaha and Audrey Richard CVPR Daily: Thursday