Seminar at Telecom ParisTech, LTCI
Abstract: In this talk, I will speak about how machine learning models can be used for visual scene understanding. Precisely, various neural network approaches will be explored and applied to practical use cases from 2D to 3D, ranging from remote sensing and standard-viewpoint image understanding to point cloud classification. We will see how to build deep network architectures which adapt to multimodal, heterogeneous data and multiple tasks. Based on the gain in performances and accuracy of the last years, I will show how it is now possible to semantize the world that surrounds us.