Referent: Dr. Loic Landrieu, Institut National de l’Information Géographique et Forestière (IGN)
Veranstalter: Martin Burger
We propose a new supervized learning framework for oversegmenting 3D point clouds into superpoints.
We cast this problem as learning deep embeddings of the local geometry and radiometry of 3D points, such that the border of objects presents high contrasts. The embeddings are computed using a lightweight neural network operating on the points’ local neighborhood. Finally, we formulate superpoint oversegmentation as a graph-structured optimization problem with respect to the learned embeddings.