The Contextual Loss

Hey All,

We are excited to announce our next meetup

This time our lecturer would be Roey Mechrez a PhD student at Technion. Roey will present his recent work "The Contextual Loss for Image Transformation with Non-Aligned Data" (https://arxiv.org/abs/1803.02077)

Our sponsor this meetup would be LEO Pharma's Innovation Lab. LEO ILab, partnering with eHealth Ventures introduce an innovation challenge[Over 3M NIS in prizes] to explore how passive data might be used to develop predictive solutions to help patients, better understand chronic diseases and create better healthcare. Learn more here: http://www.ehealthventures.com/call-for-innovation-passive-data/

Abstract:
Feed-forward CNNs trained for image transformation problems rely on loss functions that measure the similarity between the generated image and a target image. Most of the common loss functions assume that these images are spatially aligned and compare pixels at corresponding locations. However, for many tasks, the aligned training pairs of images will not be available. I will present an alternative loss function that does not require alignment, thus providing an effective and simple solution for a new space of problems. I will further show that training with our loss helps to maintain natural image statistics, a crucial factor in restoration and generation of realistic looking images.

About Roey:
Roey Mechrez is currently pursuing his Ph.D. at the Department of Electrical Engineering at the Technion. He works at the Computer Graphics & Multimedia lab under the supervision of Prof. Lihi Zelnik-Manor. He received the B.Sc. and M.Sc. degrees in Biomedical Engineering (both cum laude) from Tel-Aviv University in 2015. His research interests are in the areas of computer vision, machine learning and image processing. More specifically he is interested in photorealistic image synthesis and manipulation, image editing and image similarity. Link to Roey's web page: http://cgm.technion.ac.il/people/Roey/