NTIRE 2023 图像阴影去除挑战赛技术报告:IIM_TTI 团队
计算机视觉与模式识别
2024-03-18 v2
摘要
在本文中,我们为准备 NTIRE2023 阴影去除挑战赛 [1] 对 ShadowFormer 进行了分析和讨论,实施了五项关键改进:图像对齐、引入感知质量损失函数、阴影检测的半自动标注、阴影检测与去除的联合学习,以及引入用于阴影去除的新数据增强技术“CutShadow”。我们的方法在 LPIPS 中取得了 0.196 分(19 支队伍中排名第 3),在平均主观评分(MOS)中取得了 7.44 分(19 支队伍中排名第 4)。
引用
@article{arxiv.2403.08995,
title = {NTIRE 2023 Image Shadow Removal Challenge Technical Report: Team IIM_TTI},
author = {Yuki Kondo and Riku Miyata and Fuma Yasue and Taito Naruki and Norimichi Ukita},
journal= {arXiv preprint arXiv:2403.08995},
year = {2024}
}
备注
This version is a brief technical report submitted to the organizers, and there are still some points to be added; please wait for updates until May 2024. The code can be found here (https://github.com/Yuki-11/NTIRE2023_ShadowRemoval_IIM_TTI)