Colloquium Talk by Caner Harzibas from TU Munich
Please be advised that Caner Harzibas from the TU Munich is going to hold a colloquium talk on Feb. 16th 2017 in Room H-A 7117 at 14:00 hrs. The title of the talk is: Delving Deep into Computer Vision
Computer Vision has still major tasks to be addressed, e.g. classification, detection, segmentation and ill-posed inverse problems. Recent studies in deep learning show an outstanding performance in these tasks. Deep learning is a data-driven machine learning method which constructs hierarchical representations of data with multiple levels of abstraction. In this talk, I will discuss the major advantages of deep learning methods in computer vision applications by focusing on Convolutional Neural Networks and recent works, including our latest publications on optical flow estimation, semantic segmentation using depth information, and image based localization.