The research topic “Probabilistic Inverse Graphics problem” involves solving reconstruction problems, which are particular case of nonlinear optimisation. The way we’re approaching this task is by performing search in the space of 3D shapes, representing various objects, to approximate given multi-modal data set, acquired from various sensors such as structured light, RGB camera, lidar, etc.
Impact of this research is an ability to create a system that makes 3D scene reconstructions with variable amount of details based on heterogeneous data and prior knowledge. A particular instance of such work is . Currently we develop a further follow up on a similar topic.
- Avetisyan, Armen and Dahnert, Manuel and Dai, Angela and Savva, Manolis and Chang, Angel X. and Niessner, Matthias. Scan2CAD: Learning CAD Model Alignment in RGB-D Scans. The IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2019