Welcome to Xiangyu Xu's Homepage!
“We are trying to prove ourselves wrong as quickly as possible, because only in that way can we find progress.” ― Richard P. Feynman
Contact me: xuxiangyu2014@gmail.com
Find my new website here.
My research interest lies at the intersection of Image Processing, Computer Vision, and Machine Learning. In particular, my work focuses on understanding different aspects of visual information in images and videos, including texture, resolution, color, motion, depth, geometry, illumination, etc., and developing algorithms to reconstruct and enhance them. My work includes BlindSR, RawSR, QVI, RSC-Net, and Texformer.
Short bio: I was a Research Fellow at Nanyang Technological University from 2020 to 2021. I did my postdoc at Carnegie Mellon University and Massachusetts Institute of Technology from 2019 to 2020. I got the Ph.D. degree in the Department of Electronic Engineering at Tsinghua University in 2018. Before this, I received my B.Eng degree in the Department of Electronic Engineering at Tsinghua University, and my B.Ec degree in the National School of Development at Peking University. I was a visiting scholar at University of California, Merced and Harvard University. I have been fortunate to work with Yu-Jin Zhang at Tsinghua, Ming-Hsuan Yang at UC Merced, Deqing Sun and Hanspeter Pfister at Harvard, Fernando De la Torre, Francesc Moreno-Noguer, and Laszlo A. Jeni at CMU, George Barbastathis at MIT, Chen Change Loy at NTU, Shuicheng Yan at Sea AI Lab. I am an Area Chair of CVPR 2023 and have served on the Senior Program Committees of AAAI 2023 and IJCAI 2021.
I am open to collaborations in a broad area of computer vision. Please drop me an email if you are interested.
Publications
Kelvin C.K. Chan, Shangchen Zhou, Xiangyu Xu, and Chen Change Loy, "Investigating Tradeoffs in Real-World Video Super-Resolution", IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2022.
[Paper]
Kelvin C.K. Chan, Shangchen Zhou, Xiangyu Xu, and Chen Change Loy, "BasicVSR++: Improving Video Super-Resolution with Enhanced Propagation and Alignment", IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2022.
[Paper]
Xiangyu Xu, Chen Change Loy, "3D Human Texture Estimation from a Single Image with Transformers", IEEE International Conference on Computer Vision (ICCV), 2021. (Oral)
[Paper, Project, ICCV Daily]
Xiangyu Xu, Hao Chen, Francesc Moreno-Noguer, Laszlo A. Jeni, and Fernando De la Torre, “3D Human Pose, Shape and Texture from Low-Resolution Images and Videos”, IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI), 2021.
Extension of RSC-Net (ECCV'20): reconstruct human textures; temporally-coherent motion from videos.
Xiangyu Xu, Yongrui Ma, Wenxiu Sun, and Ming-Hsuan Yang, "Exploiting Raw Images for Real-Scene Super-Resolution", IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI), 2020.
Extension of RawSR (CVPR'19): dense channel attention; learned guided filtering; applications in image dehazing and joint depth upsampling
Xiangyu Xu, Hao Chen, Francesc Moreno-Noguer, Laszlo A. Jeni, and Fernando De la Torre, "3D Human Shape and Pose from a Single Low-Resolution Image with Self-Supervised Learning", European Conference on Computer Vision (ECCV), 2020.
Highlight: RSC-Net -- the first 3D human estimation method for low-resolution images; self-supervised learning and contrastive learning are used to deal with weak supervision.
Xiangyu Xu, Muchen Li, Wenxiu Sun, and Ming-Hsuan Yang, "Learning Spatial and Spatio-Temporal Pixel Aggregations for Image and Video Denoising", IEEE Transactions on Image Processing (TIP), 2020.
Early version: "Learning Deformable Kernels for Image and Video Denoising", axXiv 1904.06903.
Highlight: deformable convolutional neural network (DCN) for image and video denoising; the first time DCN is extended to the spatio-temporal space
Xiangyu Xu*, Siyao Li*, Wenxiu Sun, Qian Yin, and Ming-Hsuan Yang, "Quadratic Video Interpolation", Advances in Neural Information Processing Systems (NeurIPS), 2019. (Spotlight, Winner of the AIM video interpolation challenge ICCV 2019)
Highlight: the first nonlinear video frame interpolation method
Xiangyu Xu, Deqing Sun, Sifei Liu, Wenqi Ren, Yu-Jin Zhang, Ming-Hsuan Yang, and Jian Sun, "Rendering Portraitures from Monocular Camera and Beyond", European Conference on Computer Vision (ECCV), 2018.
Highlight: the first deep learning method for monocular bokeh/defocus effect rendering
[Paper]
Yukang Gan*, Xiangyu Xu*, Wenxiu Sun, and Liang Lin, "Monocular Depth Estimation with Affinity, Vertical Pooling and Label Enhancement", European Conference on Computer Vision (ECCV), 2018.
(* indicates equal contribution)
[Paper]
Wenqi Ren*, Jingang Zhang*, Xiangyu Xu*, Lin Ma, Xiaochun Cao, Gaofeng Meng, and Wei Liu, "Deep Video Dehazing with Semantic Segmentation'', IEEE Transactions on Image Processing (TIP), 2018.
(* indicates equal contribution)
[Paper]
Xiangyu Xu, Deqing Sun, Jinshan Pan, Yu-Jin Zhang, Hanspeter Pfister, and Ming-Hsuan Yang, "Learning to Super-resolve Blurry Face and Text Images", IEEE International Conference on Computer Vision (ICCV), 2017.
Highlight: the first deep neural network for blind image super-resolution
Talks
Quadratic video interpolation, NeurIPS'19, Vancouver Canada
Visual super-resolution under diverse data forms, VALSE Webinar (online)
3D human shape and pose from a single low-resolution image, ECCV pre-conference
Professional service
Area Chair: CVPR 2023, BMVC 2023
Senior Program Committee: IJCAI 2021, AAAI 2023
Useful links
MMLab, Nanyang Technological University
3D Optical System Group, Massachusetts Institute of Technology
Robotics Insititute, Carnegie Mellon University
Vision and Learning Lab, University of California, Merced
Visual Computing Group, Harvard University
Image Engineering Lab, Tsinghua University