LabelMe

LabelMe is a desktop image annotation app for people preparing computer vision datasets. It combines manual drawing with AI assistance for outlining objects and creating labels from text. It runs on 64-bit macOS, Windows and Linux..

For manual annotation, you can draw polygons, rectangles, circles, lines and points. Image flags support classification and dataset cleaning, while video annotation extends the work beyond still images. These capabilities cover bounding box detection, semantic segmentation and instance segmentation within the same graphical app.

AI assistance works with different kinds of input. SAM and EfficientSAM turn selected points into polygons or masks; YOLO-world and SAM3 create annotations from text prompts. That gives annotators a choice between drawing shapes directly and using a model to help identify regions.

Annotations stay in local JSON files. LabelMe exports VOC datasets for semantic and instance segmentation, plus COCO datasets for instance segmentation. Its stable JSON format also lets developers read annotation files in their own code.

The interface supports predefined labels, automatic saving and label validation, with translations including Japanese, Chinese, Korean and Ukrainian. LabelMe is open source under GPL-3.0 and uses Python with Qt for its desktop interface. A standalone app is also available through labelme.io without separate Python or Qt dependencies.

Current development targets the Qt6 release line. The older Qt5 line receives critical fixes only, on a best-effort basis.