This Wednesday (27.12.17) I gave a talk in the Omek-3D academia conference about my work on 3D point cloud classification. The conference took place at Daniel Hotel in Herzliya, Israel.
In a previous Omek-3D classification forum I presented some recent work on point cloud classification using deep learning, this time I presented our (Prof. Anath Fischer, Prof. Michael Lindenbaum and me) approach on the subject. This talk included a short primer about Gaussian Mixture Models (GMMs), also available here, a primer about Fisher Vectors (FV), which will also be available here soon, and a short description of our method which includes the 3DmFV representation, 3D point cloud classification network (3DmFV-Net), and part-segmentation extension. Our approach is presented in detail in our paper “3D Point Cloud Classification and Segmentation using 3D Modified Fisher Vector Representation for Convolutional Neural Networks”. Luckily, the amazing Ronit Schneor took some pictures exactly in the right slides (where I am presenting 3DmFV and 3DmFV-Net – see my hands ? ).
The conference had many great talks. I personally liked the following talks
- “Modifying Non-Local Variations Across Multiple Views” given by the always fascinating Prof. Lihi Zelnik-Manor from Technion IIT.
- “Convolutional Neural Nets on Surfaces” given by Dr. Meirav Galun from Weizmann Institute (link to paper).
- “Point Cloud Registration using Large Triangles” given by Prof. Ilan Shimshoni From Haifa University (link to paper).
We signed off the day with an amazing view of a Mediterranean sunset.