VIDIT

VIDIT数据集由瑞士洛桑联邦理工学院创建，专注于深度图像重照明技术，旨在通过特定的光照调整实现照片增强。数据集包含390个场景，总计15,600张图像，每个场景通过40种预设光照设置捕捉，涵盖5种色温和8种光照方向。所有场景均为1024x1024像素分辨率，包含多种材料和表面。数据集通过Unreal Engine 4渲染生成，分为训练、验证和测试集，其中测试集保密用于基准测试。VIDIT数据集主要应用于图像重照明、领域适应性增强和图像编辑等领域，旨在解决光照条件下的图像处理问题。

The VIDIT dataset was developed by École Polytechnique Fédérale de Lausanne (EPFL) in Switzerland, focusing on deep image relighting techniques to achieve photo enhancement via targeted light adjustments. It contains 390 scenes with a total of 15,600 images, where each scene is captured under 40 preset lighting configurations, covering 5 color temperatures and 8 lighting directions. All scenes have a resolution of 1024×1024 pixels and feature diverse materials and surfaces. Rendered using Unreal Engine 4, the dataset is split into training, validation, and test sets, with the test set withheld for benchmarking. The VIDIT dataset is primarily applied in fields including image relighting, domain adaptation enhancement, and image editing, aiming to address image processing challenges under varying lighting conditions.

搜集汇总

背景与挑战

背景概述

VIDIT数据集是一个专注于深度图像重照明技术的数据集，由瑞士洛桑联邦理工学院创建，包含390个场景和15,600张图像，每个场景通过40种预设光照设置捕捉，涵盖多种色温和光照方向。数据集通过Unreal Engine 4渲染生成，分为训练、验证和测试集，主要应用于图像重照明、领域适应性增强和图像编辑，旨在解决光照条件下的图像处理问题。