ICCV Daily 2021 - Friday

general direction and the idea for this approach of integrating supervision sources into a self-supervised learning algorithm. I would try not to talk too much and just present the basic ideas and some of the results to see what feedback they have! ” Outside of this work, Mahmoud tells us he spends half of his time writing code and experiments and divides the rest between reading other papers – mainly because he enjoys it, but also because it is important to contextualize what he is doing – and brainstorming with his collaborators. This paper was a group effort with Mahmoud as first author. His co-authors brought their own expertise to the table, including Mathilde Caron, Armand Joulin, Piotr Bojanowski, Ishan Misra, who have all done some great work in the computer vision community; Nicolas Ballas, who has plenty of experience in deep learning in specific and representation learning; and Michael Rabbat , who is experienced in optimization, distributed optimization, and large-scale learning in general. All were essential for PAWS. The team would meet regularly over video conference to provide feedback, advice, support, and direction. “ It was really nice getting to see everyone and chat and we always talked about personal things as well as work, ” Mahmoud smiles. The PAWS code is available online, and it is open source for anyone who wants to experiment with it or build on it. To learn more about Mahmoud’s work [Paper ID 8327], you can view his pre-recorded oral presentation and PDF poster now and join his live Q&A session today [7B] at 15:00-17:00 EDT. 10 DAILY ICCV Friday Oral Presentation