多模态数据标签可视化, Multimodal annotation viewer/editor. Specialized for LLaMA-Factory.
2025/2/12 MultiAnnEditor Supports .exe for windows and .pkg for mac.
Image-Text Annotation Editor is a GUI application built with Python and PyQt5 for browsing, editing, and visualizing annotation data stored in JSON format. The tool supports:
- Image and Text Display: View images and corresponding text annotations.
- Multi-Format Image Support: The
image
field may be a string or a list; users can easily switch between images. - Bounding Box Visualization: Overlay bounding boxes in various formats (e.g.,
cxcywh
,xyxy
,xywh
and their normalized versions). If multiple bounding boxes exist, navigation buttons allow switching between them. - Search Navigation: A search bar lets you choose a key and input a value to jump to matching annotation records.
- Layout Adjustment: Resizable split views for image and text areas.
- Language Switching: Toggle between English and Chinese interfaces.
- Additional Options: Set an image prefix to recover image paths, define custom image fields, etc.
- Load JSON Annotations: Easily open and browse JSON files containing annotations.
- Edit Annotations: Modify text fields and update annotation data on the fly.
- Overlay Settings: Configure how bounding boxes are displayed via the "Overlay Options" dock.
- Search & Jump: Use the top search bar to select a field, enter a value, and click "Go" to jump to a matching record.
- Responsive UI: Drag and adjust layout areas as needed.
- Prerequisites: Python 3.6 or higher.
- Install Dependencies:
pip install PyQt5
- Clone the Repository:
git clone https://github.com/yourusername/annotation-editor.git cd annotation-editor
- Run the Application:
python annotation_editor.py
- Loading Annotations: Open a JSON file via the "File" menu.
- Editing & Navigation: Edit text fields; use the "Previous"/"Next" buttons to switch records.
- Overlay Settings: Configure bounding box display in the "Overlay Options" dock.
- Search: Use the top search bar to select a field, enter a value, and click "Go" to jump to a matching record.
- Language Switch: Use the language buttons to toggle between English and Chinese.
Contributions are welcome! Please submit issues or pull requests for improvements.
This project is licensed under the MIT License.
图片-文本标注编辑器是一个基于 Python 和 PyQt5 开发的图形界面应用,用于浏览、编辑和可视化存储在 JSON 格式中的标注数据。该软件支持:
- 图片与文本显示: 显示图片及其对应的文本标注。
- 多格式图片支持:
image
字段可以为字符串或列表,支持通过下拉框切换多图显示。 - bbox 可视化: 支持以不同格式(如
cxcywh
、xyxy
、xywh
及其归一化版本)叠加显示 bounding box。若存在多个 bbox,可通过按钮切换显示。 - 搜索跳转: 顶部搜索栏允许选择字段并输入查询值,快速跳转到匹配的标注记录。
- 布局调整: 左右区域支持拖拽调整显示比例。
- 语言切换: 支持中英文界面切换。
- 其他选项: 可设置图片前缀以补救无效的图片路径,可手动定义图片字段等。
- 加载 JSON 标注: 通过“文件”菜单打开 JSON 标注文件。
- 编辑标注: 修改文本标注并实时保存。
- 叠加设置: 在“叠加选项”区域配置 bbox 显示格式及多个 bbox 之间的切换。
- 搜索跳转: 使用顶部搜索栏,根据指定字段和值快速跳转到匹配记录。
- 响应式界面: 支持左右区域拖拽调整布局。
- 前提条件: Python 3.6 及以上版本。
- 安装依赖:
pip install PyQt5
- 克隆仓库:
git clone https://github.com/likaiucas/MultiAnnEditor.git cd annotation-editor
- 运行软件:
python annotation_editor.py
- 加载标注数据: 通过“文件 → 打开 JSON”加载标注文件。
- 编辑与翻页: 在右侧编辑区域修改标注数据,并使用“上一条/下一条”按钮切换记录。
- 叠加设置: 在“叠加选项”中配置 bbox 显示方式,如选择格式(cxcywh, xyxy, xywh 等)及切换多个 bbox。
- 搜索跳转: 顶部搜索栏允许选择字段并输入查询值,点击“跳转”按钮即可跳转到匹配记录。
- 语言切换: 使用右上角的语言按钮在中英文之间切换。
欢迎开发者提交 issue 和 pull request,共同改进此项目。
本项目采用 MIT 许可证,详情请参阅 LICENSE。
This README includes both English and Chinese sections. You can simply scroll down to view the version you need.