鲁棒语义分割 UNCV2023 挑战赛结果
计算机视觉与模式识别
2023-09-28 v1 机器学习
摘要
本文概述了在 ICCV 2023 举办的 MUAD 不确定性量化挑战赛中用于解决问题的获胜方案。该挑战赛围绕城市环境中的语义分割展开，特别关注自然对抗场景。报告展示了 19 份提交结果，众多技术受到过去几年计算机视觉与机器学习领域重要会议及期刊上提出的前沿不确定性量化方法的启发。本文档中介绍了该挑战赛，阐明其目的与目标，主要围绕增强城市场景在自然对抗条件变化下语义分割的鲁棒性。报告随后深入探讨表现最优的方案。此外，本文档旨在全面概述所有参与者所采用的多样化方案。借此，力求为读者提供更深入的见解，了解可有效处理自动驾驶与语义分割（尤其是城市环境中）固有不确定性的系列策略。
引用
@article{arxiv.2309.15478,
title = {The Robust Semantic Segmentation UNCV2023 Challenge Results},
author = {Xuanlong Yu and Yi Zuo and Zitao Wang and Xiaowen Zhang and Jiaxuan Zhao and Yuting Yang and Licheng Jiao and Rui Peng and Xinyi Wang and Junpei Zhang and Kexin Zhang and Fang Liu and Roberto Alcover-Couso and Juan C. SanMiguel and Marcos Escudero-Viñolo and Hanlin Tian and Kenta Matsui and Tianhao Wang and Fahmy Adan and Zhitong Gao and Xuming He and Quentin Bouniot and Hossein Moghaddam and Shyam Nandan Rai and Fabio Cermelli and Carlo Masone and Andrea Pilzer and Elisa Ricci and Andrei Bursuc and Arno Solin and Martin Trapp and Rui Li and Angela Yao and Wenlong Chen and Ivor Simpson and Neill D. F. Campbell and Gianni Franchi},
journal= {arXiv preprint arXiv:2309.15478},
year = {2023}
}
备注
11 pages, 4 figures, accepted at ICCV 2023 UNCV workshop