報告題目：Image Enhancement with Deep Learning
Xiangyu Xu has been a postdoc researcher at Carnegie Mellon University and Massachusetts Institute of Technology from 2019 to 2020. He received the B.E. and Ph.D. degrees in 2013 and 2018 from the Department of Electronic Engineering, Tsinghua University, Beijing, China. He was a research scientist at SenseTime, Beijing. He was a visiting researcher at University of California, Merced and Harvard University. His research interest includes image processing, computer vision, and machine learning. He regularly serves on program committees and reviews papers for major computer vision and machine learning conferences and journals. He published over 10 papers on CCF A ranked top conferences and journals, including IEEE TPAMI、NeurIPS、ICML、CVPR、ICCV.
Image enhancement is an important research field, lying at the intersection of computer vision, signal processing, and computer graphics. It aims to improve the perception quality and/or enhance the visual contents of the original image, and has wide applications in film industry, social media, video surveillance, and remote sensing. In this seminar, I will give an overview of several representative works on image enhancement using deep learning techniques. In particular, I will go deeper into two of my most recent works, including quadratic video interpolation (NeurIPS'19), and 3D human reconstruction from low-resolution input (ECCV'20).