ICCV Daily 2021 - Friday

This new workshop gives due prominence to those algorithms and methods that address computer vision problems in a classic way , in the sense that they exploit explicit or analytic models , as opposed to neural or learned ones. In particular, it aims to give focus to understanding when a traditional approach works better than a deep learning solution. For instance, in terms of generalization efforts to collect data or computational requirements. But in these modern times, hasn’t deep learning replaced everything ? “ We want to put the focus on all these cool approaches that have been overshadowed by deep learning over the years, ” Matteo responds. “ Now, you go to CVPR and ICCV, and you read deep, deep, neural, neural everywhere. But the message we would like to get across with this workshop is that you can’t make everything based on learning. Sometimes you need to also reason with your mind and design the solution yourself. To be clear, we’re not saying deep learning is the devil! I work a lot with deep learning myself. I’d say deep learning is not the only answer. We can also use everything we were used to working with before, together with deep learning when we can, or in place of deep learning in some cases. ” 16 DAILY ICCV Friday Workshop Preview Introduction to Event Detection Cameras They are co-organizers of the 1st Workshop on Traditional Computer Vision in the Age of Deep Learning (TradiCV) at ICCV 2021. They speak to us ahead of the main event on Saturday. Before we start, we must commend Federica for her Honorable Mention in the Paper Awards this week for her work, Viewing Graph Solvability via Cycle Consistency. Congratulations, Federica! Federica Arrigoni is an Assistant Professor at the University of Trento. Closing the Lo p Betwe n Vision nd L nguage “… We’re not saying deep learning is the devil!” Matteo Poggi is an Assistant Professor at the University of Bologna.