无需相机位姿和场景微调，一键修复低光照三维场景
Lumos3D: A Single-Forward Framework for Low-Light 3D Scene Restoration
- 用教师-学生蒸馏提取几何信息，无需预设相机位姿
- 单次前向传播即可恢复光照与结构，无须逐场景优化
- 适用于真实世界多视角低光图像，效果媲美专用方法
低光条件下的三维场景重建极具挑战性，现有方法大多依赖预先计算的相机位姿和针对特定场景的优化，严重限制了其在真实场景中的应用。为此，我们提出Lumos3D，一种无需位姿的单次前向三维低光场景修复框架。首先，设计跨光照蒸馏机制，冻结的教师网络以正常光照真值图像为输入，向学生模型蒸馏精确几何信息。其次，定义Lumos损失函数，提升重建三维高斯空间的恢复质量。Lumos3D仅在单一数据集上训练，推理时完全前向传播，直接从无位姿的多视角低光图像中恢复光照与结构，无需任何场景级训练或优化。在真实世界数据集上的实验表明，其恢复效果可与场景专用方法相媲美。代码已开源：https://github.com/HanzhouLiu/Lumos3D。
原文摘要 · Abstract (English)
Restoring 3D scenes with low-light conditions is challenging, and most existing methods depend on precomputed camera poses and scene-specific optimization, which greatly restricts their application to real-world scenarios. To overcome these limitations, we propose Lumos3D, a pose-free single-forward framework for 3D low-light scene restoration. First, we develop a cross-illumination distillation scheme, where a frozen teacher network takes normal-light ground truth images as input to distill accurate geometric information to the student model. Second, we define a Lumos loss to improve the restoration quality of the reconstructed 3D Gaussian space. Trained on a single dataset, Lumos3D performs inference in a purely feed-forward manner, directly restoring illumination and structure from unposed, low-light multi-view images without any per-scene training or optimization. Experiments on real-world datasets demonstrate that Lumos3D achieves competitive restoration results compared to scene-specific methods. Our codes are available at https://github.com/HanzhouLiu/Lumos3D.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。