LORIA1/InstructCLIP-InstructPix2Pix-Data
收藏资源简介:
该数据集用于训练指令引导的图像编辑模型。它基于InstructPix2Pix CLIP-filtered数据集构建,并引入了新的编辑指令。每个数据样本包含四个关键元素:source_image(原始图像)、instruction(编辑指令)、target_image(编辑后的图像)以及original_instruction(来自原始InstructPix2Pix CLIP-filtered数据集的编辑指令)。数据集旨在通过自动化数据精炼方法,提升基于对比学习的指令引导图像编辑模型的性能。
This dataset is designed for training instruction-guided image editing models. It is built upon the InstructPix2Pix CLIP-filtered dataset with the addition of new edit instructions. Each sample includes four key components: source_image (the original image), instruction (the edit instruction), target_image (the edited image), and original_instruction (the edit instruction from the original InstructPix2Pix CLIP-filtered dataset). The dataset aims to enhance the performance of instruction-guided image editing models using automated data refinement through contrastive learning.




