Seminar on "Vocabulary tree based recognition for loop closing and place recognition"
Place and date: CLA E.32, on Tuesday the 11th at 14.00
Title: Vocabulary tree based recognition for loop closing and place recognition
Abstract: In this talk I will present a highly scalable approach for image search and vision based object recognition based on a vocabulary tree. The approach is very suitable for vision based place recognition and loop closing. I will introduce the basics of this approach and give an analysis of the scalability, memory needs and speed of the method.
I will show results of a mobile robot using this method for indoor navigation and show results for outdoor place recognition.
Bio:
I did my PhD at the Institute for Computer Graphics and Vision (ICG) at the Graz University of Technology, Austria. My PhD thesis focused on piecewise planar 3D reconstruction and visual localization and map building for mobile robots. After that I was Post-Doc researcher at the University of Kentucky, Lexington and the University of North Carolina, Chapel Hill. Currently I am a Post-Doc researcher at the Computer Vision and Geometry Group at ETH Zurich.
My research interests include large scale object recognition, large scale structure from motion and vision based localization with application to mobile robots.