Text2Relight 数据集
收藏资源简介:
Text2Relight 数据集是一个用于文本引导肖像重光照的合成数据集。该数据集通过自动生成多样化的文本提示,并结合文本引导的图像生成模型,创建了包含前景和背景重光照的图像对。数据集的创建过程包括文本生成、文本驱动的光照图像生成和基于图像的重光照。该数据集主要应用于文本引导的肖像重光照任务,旨在解决现有数据集在光照特定场景下缺乏多样性和可扩展性的问题。
The Text2Relight dataset is a synthetic dataset developed for text-guided portrait relighting tasks. It automatically generates diverse text prompts and leverages text-guided image generation models to create image pairs featuring both foreground and background relighting. The construction pipeline of this dataset includes three core stages: text generation, text-driven lighting image generation, and image-based relighting. This dataset is primarily applied to text-guided portrait relighting tasks, aiming to address the shortcomings of existing datasets that lack diversity and scalability in specific lighting scenarios.




