Mauricio Delbracio, “Improving realism in natural image restoration”
Abstract: Image restoration has seen tremendous progress in recent years mostly coming hand-in-hand with the success of deep neural networks. Greater computational power, stable and accessible training frameworks and large amounts of data have enabled deep image processing models that exceed or are on par with those conceived through careful and artisan modeling. However, training deep learning models to restore images that capture the realism of natural images while being faithful to the low quality input remains an open challenge.
In this talk I will present two of our recent efforts in this direction. The first method relies on a supervised point estimation that leverages features obtained from object recognition CNNs for measuring perceptual similarities between images. To faithfully compare images we borrow ideas from Optimal Transport Theory and efficiently compare distribution of features using the sliced Wasserstein distance. The second approach builds on the recent success of denoising diffusion models to cast the restoration problem as one of sampling from the posterior distribution. Unlike existing techniques, we train a stochastic sampler that refines the output of a deterministic predictor and is capable of producing a diverse set of plausible reconstructions for a single input.
I will close the talk with a discussion on some open challenges I believe need to be addressed to see even more progress in this area.
This talk is based on joint work with: Hossein Talebi, Jay Whang, Peyman Milanfar, and many others.
Biography: Mauricio Delbracio is a Senior Research Scientist at Google Research. Before joining Google in 2019, he was an Assistant Professor at the Department of Electrical Engineering, Universidad de la República (UdelaR), Uruguay. From 2013 to 2016 he was a postdoctoral researcher with the ECE Department at Duke University. He received the B.Sc degree in electrical engineering from UdelaR, Montevideo, in 2006, and the M.Sc. and Ph.D. degrees in applied mathematics from École Normale Supérieure de Cachan (ENS-Cachan), France, in 2009 and 2013 respectively. His research interests include image and signal processing, computer graphics, computational imaging, and machine learning. His current research focuses on algorithms, data analysis and applications of machine learning to image and signal processing. In 2016 he was awarded the Early Career Prize from the Society for Industrial and Applied Mathematics (SIAM) Activity Group on Imaging Science in 2016 for his significant contributions to image processing.
Join Zoom Meeting
https://wse.zoom.us/j/99304114570
Meeting ID: 993 0411 4570
One tap mobile
+13017158592,,99304114570# US (Washington DC) 13126266799,,99304114570#
+US (Chicago)
Dial by your location
+1 301 715 8592 US (Washington DC)
+1 312 626 6799 US (Chicago)
+1 646 558 8656 US (New York)
+1 669 900 6833 US (San Jose)
+1 253 215 8782 US (Tacoma)
+1 346 248 7799 US (Houston)
Meeting ID: 993 0411 4570
Find your local number: https://wse.zoom.us/u/acPT2svkU3
Join by SIP
Join by H.323
162.255.37.11 (US West)
162.255.36.11 (US East)
115.114.131.7 (India Mumbai)
115.114.115.7 (India Hyderabad)
213.19.144.110 (Amsterdam Netherlands)
213.244.140.110 (Germany)
103.122.166.55 (Australia Sydney)
103.122.167.55 (Australia Melbourne)
149.137.40.110 (Singapore)
64.211.144.160 (Brazil)
149.137.68.253 (Mexico)
69.174.57.160 (Canada Toronto)
65.39.152.160 (Canada Vancouver)
207.226.132.110 (Japan Tokyo)
149.137.24.110 (Japan Osaka)
Meeting ID: 993 0411 4570