• Radu Timofte, ETH Zurich,
• Andrey Ignatov, ETH Zurich,
• Luc Van Gool, KU Leuven & ETH Zurich,
• Wangmeng Zuo, Harbin Institute of Technology,
• Ming-Hsuan Yang, University of California at Merced & Google,
• Kyoung Mu Lee, Seoul National University,
• Liang Lin, Sun Yat-Sen University,
• Eli Shechtman, Creative Intelligence Lab at Adobe Research,
• Kai Zhang, ETH Zurich,
• Dario Fuoli, ETH Zurich,
• Zhiwu Huang, ETH Zurich,
• Martin Danelljan, ETH Zurich,
• Shuhang Gu, ETH Zurich & University of Sydney,
• Ming-Yu Liu, NVIDIA Research,
• Seungjun Nah, Seoul National University,
• Sanghyun Son, Seoul National University,
• Jaerin Lee, Seoul National University,
• Andres Romero, ETH Zurich,
• Hannan Lu, Harbin Institute of Technology
• Ruofan Zhou, EPFL,
• Majed El Helou, EPFL,
• Sabine Süsstrunk, EPFL,
• Roey Mechrez, BeyondMinds & Technion,
• Pengxu Wei, Sun Yat-Sen University
• Evangelos Ntavelis, CSEM & ETH Zurich,
• Siavash Bigdeli, CSEM

• Cosmin Ancuti, UPT
• Siavash Bigdeli, CSEM
• Michael S. Brown, York University
• Jianrui Cai, Hong Kong Polytechnic University
• Chia-Ming Cheng, MediaTek
• Cheng-Ming Chiang, MediaTek
• Sunghyun Cho, POSTECH
• Martin Danelljan, ETH Zurich
• Chao Dong, SIAT
• Weisheng Dong, Xidian University
• Touradj Ebrahimi, EPFL
• Majed El Helou, EPFL
• Corneliu Florea, University Politechnica of Bucharest
• Dario Fuoli, ETH Zurich
• Peter Gehler, Amazon
• Bastian Goldluecke, University of Konstanz
• Shuhang Gu, ETH Zurich & University of Sydney
• Zhe Hu, Hikvision Research
• Zhiwu Huang, ETH Zurich
• Andrey Ignatov, ETH Zurich
• Phillip Isola, MIT
• In So Kweon, KAIST
• Christian Ledig, VideaHealth
• Jaerin Lee, Seoul National University
• Kyoung Mu Lee, Seoul National University
• Seungyong Lee, POSTECH
• Victor Lempitsky, Skoltech & Samsung
• Liang Lin, Sun Yat-Sen University & DarkMatter AI
• Ming-Yu Liu, NVIDIA Research
• Hannan Lu, Harbin Institute of Technology
• Vladimir Lukin, National Aerospace University
• Roey Mechrez, BeyondMinds & Technion
• Zibo Meng, OPPO
• Yusuke Monno, Tokyo Institute of Technology
• Seungjun Nah, Seoul National University
• Hajime Nagahara, Osaka University
• Vinay P. Namboodiri, IIT Kanpur
• Evangelos Ntavelis, CSEM & ETH Zurich
• Federico Perazzi, Adobe
• Wenqi Ren, Chinese Academy of Sciences
• Andres Romero, ETH Zurich
• Aline Roumy, INRIA
• Yoichi Sato, University of Tokyo
• Nicu Sebe, University of Trento
• Eli Shechtman, Creative Intelligence Lab at Adobe Research
• Gregory Slabaugh, Huawei Noah's Ark Lab
• Sanghyun Son, Seoul National University
• Sabine Süsstrunk, EPFL
• Yu-Wing Tai, HKUST
• Hugues Talbot, Université Paris Est
• Masayuki Tanaka, Tokyo Institute of Technology
• Jean-Philippe Tarel, IFSTTAR
• Radu Timofte, ETH Zurich
• George Toderici, Google
• Luc Van Gool, ETH Zurich & KU Leuven
• Ting-Chun Wang, NVIDIA
• Xintao Wang, The Chinese University of Hong Kong
• Pengxu Wei, Sun Yat-Sen University
• Feng Yang, Google
• Ming-Hsuan Yang, University of California at Merced & Google
• Ren Yang, ETH Zurich
• Wenjun Zeng, Microsoft Research
• Kai Zhang, ETH Zurich
• Richard Zhang, UC Berkeley & Adobe Research
• Ruofan Zhou, EPFL
• Jun-Yan Zhu, Adobe Research & CMU
• Wangmeng Zuo, Harbin Institute of Technology

David Bau

MIT

Title: Reflected Light and Doors in the Sky: Rewriting a GAN's Rules [slides, video]

Abstract: Modern GANs learn remarkable representations that encode knowledge about relationships in the visual world. What is the shape of a child's eyebrow? Where do trees grow? In this talk we will dissect the internals of a GAN in order to ask how it represents such rules. Then we will ask: can we build a new world by rewriting those rules? If we edit a few weights of the GAN, how can we create a new model that goes beyond the training data, to synthesize a modified world that follows new rules that we design?

Bio: David Bau is a PhD student at MIT studying Computer Vision with Antonio Torralba. His research focuses on the dissection, visualization, and interactive manipulation of deep networks in vision. David was previously an engineer at Google, and he is also coauthor of a widely-used textbook on numerical linear algebra. David believes that, to expand human agency, machine learning should be more transparent.

Richard Zhang

Adobe Research

Title: Style and Structure Disentanglement for Image Manipulation [slides, video]

Abstract: A fundamental challenge of image manipulation problems is separating “style” from “structure”. We explore this in two settings: (a) unpaired image translation, where two image collections are given and (b) the purely unsupervised setting, where an unlabeled image collection is given. In unpaired image translation, the domain labels (for example, horses and zebras) are provided and indicate “style”. A successful model must then discover the structural correspondence between the two domains. In other words, each patch in the output should faithfully reflect the content of the corresponding patch in the input, independent of domain. We propose a straightforward method for doing so -- maximizing mutual information between the two, using a framework based on contrastive learning. Our framework, Contrastive Unpaired Translation, does not rely on any prespecified similarity function (such as L1 or perceptual loss) and enables one-sided translation, improving quality and reducing training time.
Next, we explore the purely unsupervised setting, where an unlabeled image collection is given, and structure and style must both be discovered. While generative models have become increasingly effective at producing realistic images from such a collection, adapting such models for controllable manipulation of existing images remains challenging. We propose the Swapping Autoencoder, which is designed specifically for image manipulation, rather than random sampling. The key idea is to encode an image with two independent components and enforce that any swapped combination maps to a realistic image. In particular, we enforce one component to encode co-occurrent patch statistics across different parts of an image, corresponding to its “style”. Such a distentangled representation allows us to flexibly manipulate real images in various ways, including texture swapping, local and global editing, and latent code vector arithmetic.

Bio: Richard Zhang is a Research Scientist at Adobe Research, with interests in computer vision, deep learning, machine learning, and graphics. He obtained his PhD in EECS, advised by Professor Alexei A. Efros, at UC Berkeley in 2018. He graduated summa cum laude with BS and MEng degrees from Cornell University in ECE. He is a recipient of the 2017 Adobe Research Fellowship.

Ravi Ramamoorthi

University of California, San Diego

Title: Light Fields and View Synthesis from Sparse Images: Revisiting Image-Based Rendering [slides, video]

Abstract: Light Fields and sparse view synthesis have the potential to enable easy acquisition of objects and scenes for virtual reality, display, and combining with computer graphics imagery. However, light fields are expensive to acquire and often involve a tradeoff of spatial and angular resolution. Moreover, View Synthesis reprises the classic problem of image-based rendering, which is challenging for situations involving occlusions or specularity.
In this talk, we present our work over the last 4 years on addressing light field reconstruction and view synthesis from sparse images using a physically-motivated deep learning approach. We start by describing local light field fusion, a method to achieve guaranteed sampling bounds for light fields from sparse images. We then discuss our efforts chronologically, starting with light fields from only the four corner views, light fields from a single legacy photograph and light field video. We close by briefly reviewing two of our ECCV contributions involving volumetric representations for view synthesis and relighting: including neural radiance fields and deep reflectance volumes.

Bio: Ravi Ramamoorthi is the Ronald L. Graham professor of Computer Science at the University of California, San Diego, and founding Director of the UC San Diego Center for Visual Computing. Prof. Ramamoorthi is an author of more than 150 refereed publications in computer vision and computer graphics, and has played a key role in building multi-faculty research groups that have been recognized as leaders in computer graphics and computer vision at Columbia, Berkeley and UCSD. His research has been recognized with a half-dozen early career awards, including the ACM SIGGRAPH Significant New Researcher Award in computer graphics in 2007, and the Presidential Early Career Award for Scientists and Engineers (PECASE) for his work in physics-based computer vision in 2008. He was elevated to IEEE and ACM Fellow in 2017, and inducted into the SIGGRAPH Academy in 2019.

Peyman Milanfar

Google

Title: Modern Computational Photography: Burst Imaging and Handheld Multi-frame Super-Resolution [slides, video]

Abstract: Modern algorithmic and computing advances, including machine learning, have changed the rules of photography, bringing to it new modes of capture, post-processing, storage, and sharing. In this talk, I’ll give a brief overview of recent progress in computational photography and describe some of the key advances of this technology, including burst photography and multi-frame super-resolution.

Bio: Dr. Peyman Milanfar is a Principal Scientist / Director at Google Research, where he leads the Computational Imaging team. Prior to this, he was a Professor of Electrical Engineering at UC Santa Cruz from 1999-2014. He was Associate Dean for Research at the School of Engineering from 2010-12. From 2012-2014 he was on leave at Google-x, where he helped develop the imaging pipeline for Google Glass. Most recently, Peyman’s team at Google developed the digital zoom pipeline for the Pixel phones, which includes the multi-frame super-resolution (Super Res Zoom) pipeline, and the RAISR upscaling algorithm. In addition, the Night Sight mode on Pixel 3 uses our Super Res Zoom technology to merge images (whether you zoom or not) for vivid shots in low light, including astrophotography. Peyman received his undergraduate education in electrical engineering and mathematics from the University of California, Berkeley, and the MS and PhD degrees in electrical engineering from the Massachusetts Institute of Technology. He founded MotionDSP, which was acquired by Cubic Inc. (NYSE:CUB). Peyman has been keynote speaker at numerous technical conferences including Picture Coding Symposium (PCS), SIAM Imaging Sciences, SPIE, and the International Conference on Multimedia (ICME). Along with his students, he has won several best paper awards from the IEEE Signal Processing Society.
He is a Distinguished Lecturer of the IEEE Signal Processing Society, and a Fellow of the IEEE “for contributions to inverse problems and super-resolution in imaging.”