HumanSignal/labelImgArchived
LabelImg
labelImg is a computer vision labeling tool and image bounding box annotator used to create training datasets for machine learning models. It functions as a desktop utility for drawing rectangular labels on images and saving object coordinates and class names in common machine learning formats.
The tool is specifically designed to generate and edit PascalVOC formatted XML files and create image labels in the text-based format required by YOLO object detection pipelines.
The software covers object detection annotation and training data preparation, including the ability to manage label categories with distinct colors and tag difficult examples for quality verification. It provides capabilities for multi-format annotation export and the conversion of labels between XML, text, and CSV formats.
Features
- Dataset Labeling Interfaces - Provides a graphical interface for drawing bounding boxes on images to create machine learning datasets.
- Annotation Format Exporters - Transforms annotation data into industry-standard XML and text schemas for compatibility with training pipelines.
- Bounding Box Visualizers - Overlays saved label coordinates as visual bounding boxes on images for review and editing.
- Annotation - Identifies and categorizes specific objects within images using standard label formats like XML or CSV.
- Computer Vision Tools - Acts as an interactive software interface for labeling and preparing visual datasets for model training.
- Training Dataset Preparation - Creates and exports structured label files to ensure compatibility with machine learning training pipelines.
- Bounding Box Interfaces - Provides a specialized graphical user interface for creating rectangular region annotations on images.
- Label Definition Managers - Includes an interface for managing label categories, object classes, and assigning distinct colors to ensure consistent dataset annotation.
- Annotation XML Storage - Persists image labels as structured XML files to maintain coordinate data and class names.
- Multi-Format Data Exports - Saves labeled data in multiple industry-standard formats to ensure compatibility with various training pipelines.
- Pixel Coordinate Mappings - Implements pixel-level mapping of mouse interactions to image coordinates for precise bounding box definition.
- YOLO Dataset Creation - Provides a visual editor for creating image labels in the text-based format required by YOLO pipelines.
- Data Curation - Enables reviewing and verifying annotated images to filter out difficult or low-confidence samples.
- PascalVOC XML Tooling - Specifically designed to generate and edit PascalVOC formatted XML files for object detection.
- Class List Configurations - Allows loading a predefined text file of categories to constrain user input and ensure labeling consistency.
- Label Representation Conversion - Transforms image labels between XML, text, and CSV formats for use in cloud training platforms.
- Dataset Iterators - Provides file-system based navigation to iterate through images and labels for batch annotation.
- Qt Framework Integrations - Utilizes the Qt framework to build a responsive desktop graphical interface for image annotation.
- Labeled Example Persisters - Automatically saves labeled training data to files to prevent data loss during the annotation process.
- Annotation Quality Verifications - Enables flagging images as verified and marking objects as difficult to ensure the quality of training data.
- Data Label Difficulty Markers - Provides the ability to mark specific annotations as difficult to signal low confidence samples during dataset curation.
