BleachNick/UltraEdit_Region_Based_100k

--- dataset_info: features: - name: clip_sim_source dtype: float64 - name: clip_sim_target dtype: float64 - name: clip_sim_dir dtype: float64 - name: clip_sim_image dtype: float64 - name: dinov2_sim dtype: float64 - name: ssim dtype: float64 - name: source_caption dtype: string - name: target_caption dtype: string - name: idx dtype: int64 - name: edit_prompt dtype: string - name: edit_object dtype: string - name: source_image dtype: image - name: edited_image dtype: image - name: mask_image dtype: image splits: - name: RegionBase num_bytes: 46307944017.231 num_examples: 108179 download_size: 46020450487 dataset_size: 46307944017.231 configs: - config_name: default data_files: - split: RegionBase path: data/RegionBase-* license: cc-by-4.0 task_categories: - text-to-image language: - en size_categories: - 100K<n<1M --- ## Bibtex citation ```bibtex @misc{zhao2024ultraeditinstructionbasedfinegrainedimage, title={UltraEdit: Instruction-based Fine-Grained Image Editing at Scale}, author={Haozhe Zhao and Xiaojian Ma and Liang Chen and Shuzheng Si and Rujie Wu and Kaikai An and Peiyu Yu and Minjia Zhang and Qing Li and Baobao Chang}, year={2024}, eprint={2407.05282}, archivePrefix={arXiv}, primaryClass={cs.CV}, url={https://arxiv.org/abs/2407.05282}, } ```The dataset, named UltraEdit, focuses on instruction-based fine-grained image editing. It contains 108,179 samples, each with multiple features such as clip_sim_source, clip_sim_target, clip_sim_dir, clip_sim_image, dinov2_sim, ssim, source_caption, target_caption, idx, edit_prompt, edit_object, source_image, edited_image, and mask_image. The datasets task category is text-to-image, suitable for English environments, and its size category is 100K<n<1M. The dataset is licensed under cc-by-4.0.

原始信息汇总

数据集概述

数据集信息

特征

• clip_sim_source: 数据类型为 float64
• clip_sim_target: 数据类型为 float64
• clip_sim_dir: 数据类型为 float64
• clip_sim_image: 数据类型为 float64
• dinov2_sim: 数据类型为 float64
• ssim: 数据类型为 float64
• source_caption: 数据类型为 string
• target_caption: 数据类型为 string
• idx: 数据类型为 int64
• edit_prompt: 数据类型为 string
• edit_object: 数据类型为 string
• source_image: 数据类型为 image
• edited_image: 数据类型为 image
• mask_image: 数据类型为 image

数据分割

• RegionBase: 包含 108179 个样本，总大小为 46307944017.231 字节

数据集大小

• 下载大小: 46020450487 字节
• 数据集大小: 46307944017.231 字节

配置

• default: 数据文件路径为 data/RegionBase-*

许可证

• cc-by-4.0

任务类别

• text-to-image

语言

• en

数据集规模

• 100K<n<1M

搜集汇总

背景与挑战

背景概述

该数据集包含约108k个样本，用于基于指令的细粒度图像编辑，每个样本提供源图像、编辑后图像、掩码、编辑提示和编辑对象，适用于文本到图像生成任务。