- Home
- AI Image Annotation Tools
- LabelMe
LabelMe
Cambridge, Massachusetts, United States
What it is
LabelMe is a free web-based image annotation tool for researchers who label images and share annotations. It supports polygon, scribble, and mask annotations; users can create or edit polygons, rename objects, adjust control points, or delete them. URL options provide a Mechanical Turk mode and navigation to another image in a folder, a random image, or the next image in a collection. A command-line script can create collections from all images or from a specified folder. Annotations are saved as XML in an Annotations folder, with downloads available through a MATLAB toolbox or tar files. The toolbox can retrieve a dataset or access online images and annotations directly, though the page says online access may be slow. The source is intended for installation on a web server and use through a browser, and Docker files are included. Setup calls for Apache, server-side includes, and Perl/CGI support. Instructions specify .jpg images and simple alphanumeric folder and file names; the server also needs image-read and annotation-folder write permissions.
For and against
For
- Free tool with polygon, scribble, and mask annotations
- Annotations are saved as XML
- MATLAB toolbox or tar-file downloads are described
- Docker files are included in the repository
Against
- Server setup requires Apache and Perl/CGI support
- Quick start specifies .jpg files and simple names
- Server needs writable annotation and temporary folders
- Online dataset access may be slow
The verdict
LabelMe provides several annotation modes and XML output for image-labeling work. Its web-server requirements and image naming constraints are important to consider before setup.
Who it suits
It suits researchers and computer vision teams who need image annotations in polygon, scribble, or mask form. It is also relevant to groups prepared to install and maintain the tool on a web server.
On the record
- Annotations
- The tool saves annotations in an Annotations folder and supports scribble mode with masks and scribbles.github.com · read 7 Oct 2026
- Collections
- A command-line script can create image collections from all images or from a specified folder.github.com · read 7 Oct 2026
- Configuration
- URL options can select images and collections, set a labeling username, restrict available actions, and supply a dropdown list of object names.github.com · read 7 Oct 2026
- Data export
- Annotations are saved as XML files, and the page describes downloads through a MATLAB toolbox or tar files.people.csail.mit.edu · read 7 Oct 2026
- Dataset access
- The MATLAB toolbox can download the dataset or access online images and annotations directly; the page says online access may be slow.people.csail.mit.edu · read 7 Oct 2026
- Deployment
- The source code is intended to be installed on a web server and used through a browser; the repository includes Docker files for installation.github.com · read 7 Oct 2026
- Download
- The source code is available as a zip download or through the GitHub repository.github.com · read 7 Oct 2026
- Image annotations
- The tool supports creating and editing polygons, with URL options to rename objects, modify polygon control points, delete objects, or view polygons without editing.github.com · read 7 Oct 2026
- Image file constraint
- The quick start instructions specify .jpg images and folder and file names containing alphanumeric characters without spaces or unusual characters.github.com · read 7 Oct 2026
- Intended use
- The repository categorizes the tool under annotation and computer vision.github.com · read 7 Oct 2026
- Labeling modes
- URL options include a Mechanical Turk mode and modes for moving to the next image in a folder, a random image, or the next image in a collection.github.com · read 7 Oct 2026
- License
- The repository lists an MIT license.github.com · read 7 Oct 2026
- Logging
- The tool records its actions in annotationCache/Logs/logfile.txt, which must be writable by the server.github.com · read 7 Oct 2026
- Mechanical Turk
- The tool has a Mechanical Turk mode with options for sandbox testing, minimum polygon counts, and custom worker instructions.github.com · read 7 Oct 2026
- Output format
- Annotation file layouts can be changed by editing an XML template for a collection.github.com · read 7 Oct 2026
- Project attribution
- The repository credits the MIT Computer Science and Artificial Intelligence Laboratory.github.com · read 7 Oct 2026
- Purpose
- LabelMe is a JavaScript tool for online image labeling that lets people access the tool from anywhere and collaborate without installing or copying the image dataset to their computers.github.com · read 7 Oct 2026
- Security and permissions
- Setup instructions require readable image files and writable annotation and temporary annotation folders on the web server.github.com · read 7 Oct 2026
