[cs-talks] IVC Seminar, Zhile Ren, 12/14, 1pm, MCS 148
Harrington, Jacob Walter
jwharrin at bu.edu
Mon Dec 12 14:03:21 EST 2016
Cascaded Models for Three-Dimensional Scene Understanding
Zhile Ren, PhD Student, Brown University
Wednesday December 14th, 1pm - 2pm
Cascaded models were applied by many classic and recent computer vision applications, and proved to be extremely powerful to many visual scene understanding tasks. In this talk, I will introduce two recent works on indoor and outdoor scene understanding that utilize cascaded model. Specifically, they are:
* Three-Dimensional Object Detection and Layout Prediction using Clouds of Oriented Gradients: I will introduce new representations for 3D object detection and layout prediction in cluttered indoor scenes using RGB-D sensors, as well as modeling contextual relationships between objects and scenes using cascaded model. The work have been presented as an oral presentation at CVPR 2016.(http://cs.brown.edu/~ren/publications/cvpr2016/cvpr2016.pdf)
* Semantic Scene Flow for Autonomous Vehicles: I will introduce a cascaded model to perform scene flow prediction, and also demonstrate that semantic information can be a useful cue for outdoor scene parsing. This is an on-going work and our performance on KITTI Scene Flow dataset ranked 1st for foreground scene flow predictions using two frames.
Zhile Ren is a fourth-year PhD student in Brown University working with Professor Erik Sudderth. His research interests are computer vision with applications in 3D scene understanding and computer graphics with applications in image manipulation. He did his undergrad in mathematics at Zhejiang University in China, and also worked as research interns in TTIC, Microsoft research, and NVIDIA.
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the cs-talks