报告题目：Scene Understanding with Deep Learning from Videos

报告人：谷东兵，School of Computer Science and Electronic Engineering, University of Essex

报告时间：2022年6月21日16:00-17:00

内容简介：

One of the key challenges in autonomous systems is the environment perception. SLAM techniques provide viable solutions to optimise the estimates of mobile platform pose and environment structures.  With the great success of deep learning networks in image classification, the research focus of SLAM techniques shifts from low level sparse structure features to high level dense semantic segmentation. This talk provides an overview on SLAM algorithms in two separate frameworks: geometry and deep learning, and attempts to establish the underlying link between these two frameworks. The talk demonstrates how the geometry constraints can be exploited in line with the definition of cost functions for unsupervised learning.   The talk also presents some recent advances on 3D dense perception tasks, including depth estimation, optical flow estimation, semantic segmentation, dynamic object detection together with pose estimation.

报告人简介：

Dongbing Gu is a Professor in the School of Computer Science and Electronic Engineering, University of Essex. He received the B.Sc. and M.Sc. degrees in automatic control from Beijing Institute of Technology, Beijing, China and the Ph.D. degree in robotics from University of Essex, UK.  He was an Academic Visiting Scholar with the Department of Engineering Science, University of Oxford, Oxford, UK, from October 1996 to October 1997. His research expertise is in the field of robotics, autonomous systems, and machine learning. He has published more than 200 papers in international conferences and journals. His research has been supported by UK Research Councils (EPSRC), InnovateUK, Royal Academy of Science, European Commission, and industry. Currently he is the Specialty Chief Editor of Frontiers in Robotics and AI: Field Robotics and an Editorial Board Member of Cognitive Computation. He served the organisation committee and programme committee for many international conferences.