universal-data-tool

Why we included this project

Universal Data Tool covers the labeling jobs most ML projects need, from image segmentation and classification to text and entity recognition, all in one interface. You can work in the browser, install the desktop app (Windows, Linux, and Mac are supported), or run it on your own infrastructure, so it serves a solo annotator and a small team coordinating labels alike. Labels are stored in an open .udt.json or .udt.csv format instead of a proprietary export, which keeps datasets portable between tools. There is also a JavaScript library for embedding the annotator directly into your own web pages or data pipelines, handy when labeling has to live inside an existing product. Teams that want lightweight labeling collaboration without adopting a heavier commercial platform will find it a practical fit.