Computer Vision News - December 2022

22 Congrats, Doctor! Have you ever wondered if you could do more with generative models? The past few years have seen an exponential increase of papers related to image generation . However, it became clear that those systems require plenty of data to be trained efficiently, while not offering controllable outputs that may be necessary for practical applications. In his PhD thesis, Fabio deals with these issues, by proposing improvements to the capabilities of image translation networks . First, he tackles training with data scarcity , focusing on low- shot systems that can be trained by just looking at a few examples. To achieve this goal, he proposes to bring back humans in the loop, by exploiting domain Fabio Pizzati has recently completed his PhD at the ASTRA team of Inria Paris. Under the supervision of Raoul de Charette, he developed a thesis on generative models, with many applications for autonomous driving in complex conditions, such as adverse weather and rare scenarios. Currently, Fabio is committed to pursuing an academic career and has recently accepted a postdoc position in Oxford with Phil Torr. Congrats, Doctor Fabio!