SEED-Data-Edit

SEED-Data-Edit是由腾讯AI Lab和ARC Lab联合创建的混合数据集，专注于指令引导的图像编辑。该数据集包含三种类型的数据：自动化管道生成的高质量编辑数据、从互联网收集的真实世界场景数据以及由Photoshop专家标注的多轮编辑数据。数据集旨在通过大规模、多样化的图像编辑样本，促进语言引导的图像编辑模型的训练。SEED-Data-Edit不仅支持单次编辑，还支持多轮迭代编辑，模拟真实世界中的编辑过程，适用于提升图像编辑技术的实用性和精确性。

SEED-Data-Edit is a hybrid dataset jointly created by Tencent AI Lab and ARC Lab, focusing on instruction-guided image editing. This dataset includes three types of data: high-quality edited data generated via automated pipelines, real-world scene data collected from the Internet, and multi-round edited data annotated by Photoshop experts. The dataset aims to facilitate the training of language-guided image editing models through large-scale and diverse image editing samples. SEED-Data-Edit supports not only single-turn editing but also multi-round iterative editing, simulating real-world editing processes, and is suitable for enhancing the practicality and accuracy of image editing technologies.