Get the Picture! Winning Paper at ICCV

Tamar Rott Shaham, a doctoral student at the Viterbi Faculty of Electrical Engineering at the Technion – Israel Institute of Technology, won the Best Paper Award (Marr Prize) at the International Conference on Computer Vision (ICCV) together with her supervisor, Prof. Tomer Michaeli and Dr. Tali Dekel from Google Research. 

Prof. Tomer Michaeli and Tamar Rott Shaham with the chairs of the conference Prof. Svetlana Lazebnik and Prof. Kyoung Mu Lee

ICCV is one of the three most important conferences in the field of computer vision, and the winning article was selected from more than 4,000 competing submissions.  

The winning paper described a new deep learning methodology developed by the Technion team. The researchers developed an algorithm that automatically generates “invented” images based on only a single picture example – as opposed to the vast pool of images, on which current methods are based.  The field of Deep Learning traditionally involves training a neural network based on a huge collection of samples. But here, the researchers present an innovative technique for training a generative neural network, with a training set containing only one picture.

Using the trained model, new image variations can be created that contain semantic data similar to the given image. In addition, the model can perform a variety of tasks such as editing the image, turning a painting into a realistic image and even creating a short video.