LabelMe vs Prodigy
LabelMe
labelme.csail.mit.eduProdigy
prodigy.aiSide by side
| Figure | LabelMe | Prodigy |
|---|---|---|
| Price from | — | — |
| Free plan | — | — |
| Free trial | — | — |
| Open source | — | — |
| Web | Yes | Yes |
| Windows | Yes | Yes |
| macOS | — | Yes |
| Linux | Yes | Yes |
| API | — | Yes |
| Self-hosted | Yes | Yes |
| Plans published | 0 | 2 |
| Spec score · AI Image Annotation Tools | 7.5 (#17) | 8.4 (#6) |
| 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. | The standard Python web server and application can be deployed on any cloud provider or entirely on-premise. |
| Founded | 2005 | 2016 |
| Headquarters | Cambridge, Massachusetts, United States | Berlin, Germany |
Plans
In their own words
LabelMe
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 and edit polygons, rename objects, adjust polygon control points, delete objects, or view polygons without editing them. URL options support Mechanical Turk workflows and navigation to another image in a folder, a random image, or the next image in a collection. A command-line script can build collections from all images or a specified folder. Annotations are saved as XML in an Annotations folder, with downloads described through a MATLAB toolbox or tar files. The MATLAB toolbox can retrieve a dataset or access online images and annotations directly, though online access may be slow. The source is intended for installation on a web server and browser use; Docker files are included. Setup requires Apache, server-side includes, and Perl/CGI support. Quick start instructions specify .jpg images and simple alphanumeric folder and file names. The web server needs readable images and writable annotation and temporary folders.
Prodigy
Prodigy is a downloadable developer tool and Python library for creating training and evaluation data for machine-learning models. Its built-in recipes cover named entity recognition, text classification, computer vision, audio and video annotation, transcription, and model training and evaluation. A customizable web application offers 23 interfaces for work with text, images, audio, video, relations, multiple choice, HTML and conflict resolution. Users can define workflows with Python recipes and create interfaces using HTML, CSS and JavaScript. Prodigy supports spaCy, plugins for Hugging Face models and integrations with major LLM API providers. Data can be stored in SQLite, MySQL or PostgreSQL and exported to JSON. It runs on the user's own machines without connecting to the maker's or third-party servers, including on air-gapped machines. The standard Python web server and app can be deployed in the cloud or on-premise. Personal licensing costs $390, paid once, excluding tax; Company licensing costs $490, paid once per seat in packs of five, excluding tax.