Maciej Halber will present his Pre FPO on Monday, January 28, 2019 at 11am in CS 301.

Maciej Halber will present his Pre FPO on Monday, January 28, 2019 at 11am in CS 301. The members of his committee members are as follows: Thomas Funkhouser (adviser), Adam Finkelstein, Olga Russakovsky, Matthias Niessner (Technical University of Munich), and Szymon Rusinkiewicz Everyone is invited to attend his talk. The talk title and abstract follow below: Title: RGBD Pipeline for 3D Indoor Scene Understanding Abstract: RGBD cameras, devices capable of capturing both color and distance information have became a standard data acquisition device for many computer vision tasks. While providing especially valuable source of data for applications like 3D reconstruction and understanding, pipelines able to solve such problems are complex systems and come with a wide range of new challenges. In this talk, I will present my work on improving several stages of such RGBD pipeline. First, I discuss issues that arise when capturing the data with a commodity RGBD camera, alongside a calibration method to improve it. Second, I describe a method for robustly recovering camera poses from a RGBD video sequence, where I introduce a novel optimization strategy which allows for effective incorporation of man-made priors. Finally, I discuss an algorithm for transferring information between 3D reconstructions of indoor scenes captured at different times, with applications in instance and semantic segmentation.
participants (1)
-
Nicki Gotsis