UB-CVML-Group/PIE_Bench_pp
收藏资源简介:
--- license: cc-by-sa-4.0 dataset_info: - config_name: 0_random_140 features: - name: image dtype: image - name: id dtype: string - name: source_prompt dtype: string - name: target_prompt dtype: string - name: edit_action dtype: string - name: aspect_mapping dtype: string - name: blended_words dtype: string - name: mask dtype: string splits: - name: V1 num_bytes: 15594175.0 num_examples: 140 download_size: 15528907 dataset_size: 15594175.0 - config_name: 1_change_object_80 features: - name: image dtype: image - name: id dtype: string - name: source_prompt dtype: string - name: target_prompt dtype: string - name: edit_action dtype: string - name: aspect_mapping dtype: string - name: blended_words dtype: string - name: mask dtype: string splits: - name: V1 num_bytes: 8188672.0 num_examples: 80 download_size: 8174209 dataset_size: 8188672.0 - config_name: 2_add_object_80 features: - name: image dtype: image - name: id dtype: string - name: source_prompt dtype: string - name: target_prompt dtype: string - name: edit_action dtype: string - name: aspect_mapping dtype: string - name: blended_words dtype: string - name: mask dtype: string splits: - name: V1 num_bytes: 6926151.0 num_examples: 80 download_size: 6917854 dataset_size: 6926151.0 - config_name: 3_delete_object_80 features: - name: image dtype: image - name: id dtype: string - name: source_prompt dtype: string - name: target_prompt dtype: string - name: edit_action dtype: string - name: aspect_mapping dtype: string - name: blended_words dtype: string - name: mask dtype: string splits: - name: V1 num_bytes: 7513741.0 num_examples: 80 download_size: 7382006 dataset_size: 7513741.0 - config_name: 4_change_attribute_content_40 features: - name: image dtype: image - name: id dtype: string - name: source_prompt dtype: string - name: target_prompt dtype: string - name: edit_action dtype: string - name: aspect_mapping dtype: string - name: blended_words dtype: string - name: mask dtype: string splits: - name: V1 num_bytes: 4125034.0 num_examples: 40 download_size: 4061909 dataset_size: 4125034.0 - config_name: 5_change_attribute_pose_40 features: - name: image dtype: image - name: id dtype: string - name: source_prompt dtype: string - name: target_prompt dtype: string - name: edit_action dtype: string - name: aspect_mapping dtype: string - name: blended_words dtype: string - name: mask dtype: string splits: - name: V1 num_bytes: 4217839.0 num_examples: 40 download_size: 4148577 dataset_size: 4217839.0 - config_name: 6_change_attribute_color_40 features: - name: image dtype: image - name: id dtype: string - name: source_prompt dtype: string - name: target_prompt dtype: string - name: edit_action dtype: string - name: aspect_mapping dtype: string - name: blended_words dtype: string - name: mask dtype: string splits: - name: V1 num_bytes: 4274823.0 num_examples: 40 download_size: 4263550 dataset_size: 4274823.0 - config_name: 7_change_attribute_material_40 features: - name: image dtype: image - name: id dtype: string - name: source_prompt dtype: string - name: target_prompt dtype: string - name: edit_action dtype: string - name: aspect_mapping dtype: string - name: blended_words dtype: string - name: mask dtype: string splits: - name: V1 num_bytes: 4061715.0 num_examples: 40 download_size: 4005557 dataset_size: 4061715.0 - config_name: 8_change_background_80 features: - name: image dtype: image - name: id dtype: string - name: source_prompt dtype: string - name: target_prompt dtype: string - name: edit_action dtype: string - name: aspect_mapping dtype: string - name: blended_words dtype: string - name: mask dtype: string splits: - name: V1 num_bytes: 8533140.0 num_examples: 80 download_size: 8422137 dataset_size: 8533140.0 - config_name: 9_change_style_80 features: - name: image dtype: image - name: id dtype: string - name: source_prompt dtype: string - name: target_prompt dtype: string - name: edit_action dtype: string - name: aspect_mapping dtype: string - name: blended_words dtype: string - name: mask dtype: string splits: - name: V1 num_bytes: 8698695.0 num_examples: 80 download_size: 8686937 dataset_size: 8698695.0 configs: - config_name: 0_random_140 data_files: - split: V1 path: 0_random_140/V1-* - config_name: 1_change_object_80 data_files: - split: V1 path: 1_change_object_80/V1-* - config_name: 2_add_object_80 data_files: - split: V1 path: 2_add_object_80/V1-* - config_name: 3_delete_object_80 data_files: - split: V1 path: 3_delete_object_80/V1-* - config_name: 4_change_attribute_content_40 data_files: - split: V1 path: 4_change_attribute_content_40/V1-* - config_name: 5_change_attribute_pose_40 data_files: - split: V1 path: 5_change_attribute_pose_40/V1-* - config_name: 6_change_attribute_color_40 data_files: - split: V1 path: 6_change_attribute_color_40/V1-* - config_name: 7_change_attribute_material_40 data_files: - split: V1 path: 7_change_attribute_material_40/V1-* - config_name: 8_change_background_80 data_files: - split: V1 path: 8_change_background_80/V1-* - config_name: 9_change_style_80 data_files: - split: V1 path: 9_change_style_80/V1-* --- ## What is PIE-Bench++? PIE-Bench++ builds upon the foundation laid by the original [PIE-Bench dataset](https://cure-lab.github.io/PnPInversion) introduced by (Ju et al., 2024), designed to provide a comprehensive benchmark for multi-aspect image editing evaluation. This enhanced dataset contains 700 images and prompts across nine distinct edit categories, encompassing a wide range of manipulations: - **Object-Level Manipulations:** Additions, removals, and modifications of objects within the image. - **Attribute-Level Manipulations:** Changes in content, pose, color, and material of objects. - **Image-Level Manipulations:** Adjustments to the background and overall style of the image. While retaining the original images, the enhanced dataset features revised source prompts and editing prompts, augmented with additional metadata such as editing types and aspect mapping. This comprehensive augmentation aims to facilitate more nuanced and detailed evaluations in the domain of multi-aspect image editing. ## Data Annotation Guide ### Overview Our dataset annotations are structured to provide comprehensive information for each image, facilitating a deeper understanding of the editing process. Each annotation consists of the following key elements: - **Source Prompt:** The original description or caption of the image before any edits are made. - **Target Prompt:** The description or caption of the image after the edits are applied. - **Edit Action:** A detailed specification of the changes made to the image, including: - The position index in the source prompt where changes occur. - The type of edit applied (e.g., 1: change object, 2: add object, 3: remove object, 4: change attribute content, 5: change attribute pose, 6: change attribute color, 7: change attribute material, 8: change background, 9: change style). - The operation required to achieve the desired outcome (e.g., '+' / '-' means adding/removing words at the specified position, and 'xxx' means replacing the existing words). - **Aspect Mapping:** A mapping that connects objects undergoing editing to their respective modified attributes. This helps identify which objects are subject to editing and the specific attributes that are altered. ### Example Annotation Here is an example annotation for an image in our dataset: ```json { "000000000002": { "image_path": "0_random_140/000000000002.jpg", "source_prompt": "a cat sitting on a wooden chair", "target_prompt": "a [red] [dog] [with flowers in mouth] [standing] on a [metal] chair", "edit_action": { "red": {"position": 1, "edit_type": 6, "action": "+"}, "dog": {"position": 1, "edit_type": 1, "action": "cat"}, "with flowers in mouth": {"position": 2, "edit_type": 2, "action": "+"}, "standing": {"position": 2, "edit_type": 5, "action": "sitting"}, "metal": {"position": 5, "edit_type": 7, "action": "wooden"} }, "aspect_mapping": { "dog": ["red", "standing"], "chair": ["metal"], "flowers": [] }, "blended_words": [ "cat,dog", "chair,chair" ], "mask": "0 262144" } }
许可证:CC BY-SA 4.0 数据集信息: - 配置名称:0_random_140 特征: - 字段名:image(图像),数据类型:图像 - 字段名:id,数据类型:字符串 - 字段名:source_prompt(源提示词),数据类型:字符串 - 字段名:target_prompt(目标提示词),数据类型:字符串 - 字段名:edit_action(编辑操作),数据类型:字符串 - 字段名:aspect_mapping(属性映射),数据类型:字符串 - 字段名:blended_words(混合词汇),数据类型:字符串 - 字段名:mask(掩码),数据类型:字符串 数据集划分: - 划分名称:V1,字节数:15594175.0,样本数量:140 下载大小:15528907,数据集总大小:15594175.0 - 配置名称:1_change_object_80 特征: - 字段名:image(图像),数据类型:图像 - 字段名:id,数据类型:字符串 - 字段名:source_prompt(源提示词),数据类型:字符串 - 字段名:target_prompt(目标提示词),数据类型:字符串 - 字段名:edit_action(编辑操作),数据类型:字符串 - 字段名:aspect_mapping(属性映射),数据类型:字符串 - 字段名:blended_words(混合词汇),数据类型:字符串 - 字段名:mask(掩码),数据类型:字符串 数据集划分: - 划分名称:V1,字节数:8188672.0,样本数量:80 下载大小:8174209,数据集总大小:8188672.0 - 配置名称:2_add_object_80 特征: - 字段名:image(图像),数据类型:图像 - 字段名:id,数据类型:字符串 - 字段名:source_prompt(源提示词),数据类型:字符串 - 字段名:target_prompt(目标提示词),数据类型:字符串 - 字段名:edit_action(编辑操作),数据类型:字符串 - 字段名:aspect_mapping(属性映射),数据类型:字符串 - 字段名:blended_words(混合词汇),数据类型:字符串 - 字段名:mask(掩码),数据类型:字符串 数据集划分: - 划分名称:V1,字节数:6926151.0,样本数量:80 下载大小:6917854,数据集总大小:6926151.0 - 配置名称:3_delete_object_80 特征: - 字段名:image(图像),数据类型:图像 - 字段名:id,数据类型:字符串 - 字段名:source_prompt(源提示词),数据类型:字符串 - 字段名:target_prompt(目标提示词),数据类型:字符串 - 字段名:edit_action(编辑操作),数据类型:字符串 - 字段名:aspect_mapping(属性映射),数据类型:字符串 - 字段名:blended_words(混合词汇),数据类型:字符串 - 字段名:mask(掩码),数据类型:字符串 数据集划分: - 划分名称:V1,字节数:7513741.0,样本数量:80 下载大小:7382006,数据集总大小:7513741.0 - 配置名称:4_change_attribute_content_40 特征: - 字段名:image(图像),数据类型:图像 - 字段名:id,数据类型:字符串 - 字段名:source_prompt(源提示词),数据类型:字符串 - 字段名:target_prompt(目标提示词),数据类型:字符串 - 字段名:edit_action(编辑操作),数据类型:字符串 - 字段名:aspect_mapping(属性映射),数据类型:字符串 - 字段名:blended_words(混合词汇),数据类型:字符串 - 字段名:mask(掩码),数据类型:字符串 数据集划分: - 划分名称:V1,字节数:4125034.0,样本数量:40 下载大小:4061909,数据集总大小:4125034.0 - 配置名称:5_change_attribute_pose_40 特征: - 字段名:image(图像),数据类型:图像 - 字段名:id,数据类型:字符串 - 字段名:source_prompt(源提示词),数据类型:字符串 - 字段名:target_prompt(目标提示词),数据类型:字符串 - 字段名:edit_action(编辑操作),数据类型:字符串 - 字段名:aspect_mapping(属性映射),数据类型:字符串 - 字段名:blended_words(混合词汇),数据类型:字符串 - 字段名:mask(掩码),数据类型:字符串 数据集划分: - 划分名称:V1,字节数:4217839.0,样本数量:40 下载大小:4148577,数据集总大小:4217839.0 - 配置名称:6_change_attribute_color_40 特征: - 字段名:image(图像),数据类型:图像 - 字段名:id,数据类型:字符串 - 字段名:source_prompt(源提示词),数据类型:字符串 - 字段名:target_prompt(目标提示词),数据类型:字符串 - 字段名:edit_action(编辑操作),数据类型:字符串 - 字段名:aspect_mapping(属性映射),数据类型:字符串 - 字段名:blended_words(混合词汇),数据类型:字符串 - 字段名:mask(掩码),数据类型:字符串 数据集划分: - 划分名称:V1,字节数:4274823.0,样本数量:40 下载大小:4263550,数据集总大小:4274823.0 - 配置名称:7_change_attribute_material_40 特征: - 字段名:image(图像),数据类型:图像 - 字段名:id,数据类型:字符串 - 字段名:source_prompt(源提示词),数据类型:字符串 - 字段名:target_prompt(目标提示词),数据类型:字符串 - 字段名:edit_action(编辑操作),数据类型:字符串 - 字段名:aspect_mapping(属性映射),数据类型:字符串 - 字段名:blended_words(混合词汇),数据类型:字符串 - 字段名:mask(掩码),数据类型:字符串 数据集划分: - 划分名称:V1,字节数:4061715.0,样本数量:40 下载大小:4005557,数据集总大小:4061715.0 - 配置名称:8_change_background_80 特征: - 字段名:image(图像),数据类型:图像 - 字段名:id,数据类型:字符串 - 字段名:source_prompt(源提示词),数据类型:字符串 - 字段名:target_prompt(目标提示词),数据类型:字符串 - 字段名:edit_action(编辑操作),数据类型:字符串 - 字段名:aspect_mapping(属性映射),数据类型:字符串 - 字段名:blended_words(混合词汇),数据类型:字符串 - 字段名:mask(掩码),数据类型:字符串 数据集划分: - 划分名称:V1,字节数:8533140.0,样本数量:80 下载大小:8422137,数据集总大小:8533140.0 - 配置名称:9_change_style_80 特征: - 字段名:image(图像),数据类型:图像 - 字段名:id,数据类型:字符串 - 字段名:source_prompt(源提示词),数据类型:字符串 - 字段名:target_prompt(目标提示词),数据类型:字符串 - 字段名:edit_action(编辑操作),数据类型:字符串 - 字段名:aspect_mapping(属性映射),数据类型:字符串 - 字段名:blended_words(混合词汇),数据类型:字符串 - 字段名:mask(掩码),数据类型:字符串 数据集划分: - 划分名称:V1,字节数:8698695.0,样本数量:80 下载大小:8686937,数据集总大小:8698695.0 配置项: - 配置名称:0_random_140 数据文件: - 划分:V1,路径:0_random_140/V1-* - 配置名称:1_change_object_80 数据文件: - 划分:V1,路径:1_change_object_80/V1-* - 配置名称:2_add_object_80 数据文件: - 划分:V1,路径:2_add_object_80/V1-* - 配置名称:3_delete_object_80 数据文件: - 划分:V1,路径:3_delete_object_80/V1-* - 配置名称:4_change_attribute_content_40 数据文件: - 划分:V1,路径:4_change_attribute_content_40/V1-* - 配置名称:5_change_attribute_pose_40 数据文件: - 划分:V1,路径:5_change_attribute_pose_40/V1-* - 配置名称:6_change_attribute_color_40 数据文件: - 划分:V1,路径:6_change_attribute_color_40/V1-* - 配置名称:7_change_attribute_material_40 数据文件: - 划分:V1,路径:7_change_attribute_material_40/V1-* - 配置名称:8_change_background_80 数据文件: - 划分:V1,路径:8_change_background_80/V1-* - 配置名称:9_change_style_80 数据文件: - 划分:V1,路径:9_change_style_80/V1-* ## 什么是PIE-Bench++? PIE-Bench++ 以Ju等人(2024)提出的原始PIE-Bench数据集(PIE-Bench dataset)为基础,旨在为多维度图像编辑评估提供全面的基准测试集。该增强型数据集涵盖9类不同编辑场景,包含700张图像与对应提示词,支持多种编辑操作类型: - **对象级编辑操作**:图像内对象的添加、移除与修改 - **属性级编辑操作**:对象的内容、姿态、颜色与材质变更 - **图像级编辑操作**:图像背景与整体风格的调整 该增强数据集保留了原始图像资源,同时优化了源提示词与编辑提示词,并补充了编辑类型、属性映射等额外元数据。此番全面的扩充旨在推动多维度图像编辑领域开展更精细、更深入的评估研究。 ## 数据标注指南 ### 概述 本数据集的标注结构旨在为每张图像提供全面的信息,以加深对图像编辑过程的理解。每份标注包含以下核心要素: - **源提示词**:未进行任何编辑前的图像原始描述或说明文本 - **目标提示词**:应用全部编辑操作后的图像描述或说明文本 - **编辑操作**:对图像所做变更的详细说明,包括: 1. 源提示词中发生变更的位置索引 2. 所应用的编辑类型(例如:1=修改对象、2=添加对象、3=移除对象、4=修改属性内容、5=修改属性姿态、6=修改属性颜色、7=修改属性材质、8=修改背景、9=修改风格) 3. 达成预期效果所需的具体操作(例如:'+' / '-' 表示在指定位置添加/移除词汇,'xxx' 表示替换现有词汇) - **属性映射**:将待编辑对象与其对应的修改属性建立关联的映射关系,可用于识别需编辑的对象以及具体被修改的属性 ### 示例标注 以下为本数据集中一张图像的标注示例: json { "000000000002": { "图像路径": "0_random_140/000000000002.jpg", "源提示词": "a cat sitting on a wooden chair", "目标提示词": "a [red] [dog] [with flowers in mouth] [standing] on a [metal] chair", "编辑操作": { "red": {"位置": 1, "编辑类型": 6, "操作": "+"}, "dog": {"位置": 1, "编辑类型": 1, "操作": "cat"}, "with flowers in mouth": {"位置": 2, "编辑类型": 2, "操作": "+"}, "standing": {"位置": 2, "编辑类型": 5, "操作": "sitting"}, "metal": {"位置": 5, "编辑类型": 7, "操作": "wooden"} }, "属性映射": { "dog": ["red", "standing"], "chair": ["metal"], "flowers": [] }, "混合词汇": [ "cat,dog", "chair,chair" ], "掩码": "0 262144" } }
数据集概述
数据集配置
-
0_random_140
- 特征:
- image: 图像
- id: 字符串
- source_prompt: 字符串
- target_prompt: 字符串
- edit_action: 字符串
- aspect_mapping: 字符串
- blended_words: 字符串
- mask: 字符串
- 分割: V1
- 大小: 15594175.0 字节
- 示例数: 140
- 下载大小: 15528907 字节
- 数据集大小: 15594175.0 字节
- 特征:
-
1_change_object_80
- 特征:
- image: 图像
- id: 字符串
- source_prompt: 字符串
- target_prompt: 字符串
- edit_action: 字符串
- aspect_mapping: 字符串
- blended_words: 字符串
- mask: 字符串
- 分割: V1
- 大小: 8188672.0 字节
- 示例数: 80
- 下载大小: 8174209 字节
- 数据集大小: 8188672.0 字节
- 特征:
-
2_add_object_80
- 特征:
- image: 图像
- id: 字符串
- source_prompt: 字符串
- target_prompt: 字符串
- edit_action: 字符串
- aspect_mapping: 字符串
- blended_words: 字符串
- mask: 字符串
- 分割: V1
- 大小: 6926151.0 字节
- 示例数: 80
- 下载大小: 6917854 字节
- 数据集大小: 6926151.0 字节
- 特征:
-
3_delete_object_80
- 特征:
- image: 图像
- id: 字符串
- source_prompt: 字符串
- target_prompt: 字符串
- edit_action: 字符串
- aspect_mapping: 字符串
- blended_words: 字符串
- mask: 字符串
- 分割: V1
- 大小: 7513741.0 字节
- 示例数: 80
- 下载大小: 7382006 字节
- 数据集大小: 7513741.0 字节
- 特征:
-
4_change_attribute_content_40
- 特征:
- image: 图像
- id: 字符串
- source_prompt: 字符串
- target_prompt: 字符串
- edit_action: 字符串
- aspect_mapping: 字符串
- blended_words: 字符串
- mask: 字符串
- 分割: V1
- 大小: 4125034.0 字节
- 示例数: 40
- 下载大小: 4061909 字节
- 数据集大小: 4125034.0 字节
- 特征:
-
5_change_attribute_pose_40
- 特征:
- image: 图像
- id: 字符串
- source_prompt: 字符串
- target_prompt: 字符串
- edit_action: 字符串
- aspect_mapping: 字符串
- blended_words: 字符串
- mask: 字符串
- 分割: V1
- 大小: 4217839.0 字节
- 示例数: 40
- 下载大小: 4148577 字节
- 数据集大小: 4217839.0 字节
- 特征:
-
6_change_attribute_color_40
- 特征:
- image: 图像
- id: 字符串
- source_prompt: 字符串
- target_prompt: 字符串
- edit_action: 字符串
- aspect_mapping: 字符串
- blended_words: 字符串
- mask: 字符串
- 分割: V1
- 大小: 4274823.0 字节
- 示例数: 40
- 下载大小: 4263550 字节
- 数据集大小: 4274823.0 字节
- 特征:
-
7_change_attribute_material_40
- 特征:
- image: 图像
- id: 字符串
- source_prompt: 字符串
- target_prompt: 字符串
- edit_action: 字符串
- aspect_mapping: 字符串
- blended_words: 字符串
- mask: 字符串
- 分割: V1
- 大小: 4061715.0 字节
- 示例数: 40
- 下载大小: 4005557 字节
- 数据集大小: 4061715.0 字节
- 特征:
-
8_change_background_80
- 特征:
- image: 图像
- id: 字符串
- source_prompt: 字符串
- target_prompt: 字符串
- edit_action: 字符串
- aspect_mapping: 字符串
- blended_words: 字符串
- mask: 字符串
- 分割: V1
- 大小: 8533140.0 字节
- 示例数: 80
- 下载大小: 8422137 字节
- 数据集大小: 8533140.0 字节
- 特征:
-
9_change_style_80
- 特征:
- image: 图像
- id: 字符串
- source_prompt: 字符串
- target_prompt: 字符串
- edit_action: 字符串
- aspect_mapping: 字符串
- blended_words: 字符串
- mask: 字符串
- 分割: V1
- 大小: 8698695.0 字节
- 示例数: 80
- 下载大小: 8686937 字节
- 数据集大小: 8698695.0 字节
- 特征:
许可证
- 类型: cc-by-sa-4.0




