Domain Adaptation for Semantic Segmentation in Real-World Surveillance and Autonomous Cars скачать в хорошем качестве

Domain Adaptation for Semantic Segmentation in Real-World Surveillance and Autonomous Cars

In this video, we present our latest paper: “Mixture Domain Adaptation to Improve Semantic Segmentation in Real-World Surveillance” published at the WACV 2023 workshop on Real-World Surveillance. We mathematically prove the exact solution to the mixture domain adaptation problem that can adapt on the fly from several source models. Our method can be used for many tasks, including semantic segmentation for real-world surveillance and autonomous cars. 00:00​​ Introduction 00:30 Motivation and synthetic example 05:02 Our method ​​06:44 Experiments 08:43​ Conclusion Authors: Sébastien Piérard, Anthony Cioppa, Anaïs Halin, Renaud Vandeghen, Maxime Zanella, Benoît Macq, Saïd Mahmoudi, Marc Van Droogenbroeck Link to the paper: https://arxiv.org/abs/2211.10119 Code to get started with the method: https://github.com/rvandeghen/MDA Don’t hesitate to contact us if you have any questions about the paper or the code! To see more of our work, subscribe to our YouTube channel and other social media! We’ll keep posting our work regularly. Subscribe to Acad Research:    / acadresearch   Twitter:   / anthony_cioppa​​​​   Video Credits ------- Background videos from Videezy.com and Pexels.com. ------- Music credits ------- Juno in The Space Maze - Loopop Classic - Joakim Karud ------- Presentation written by Sébastien Piérard, Anthony Cioppa, Anais Halin, and Renaud Vandeghen. Video produced and edited by Anthony Cioppa and Renaud Vandeghen.