LabelMe

LabelMe is a desktop image annotation app for people preparing computer vision datasets. It combines manual drawing with model assistance for labeling objects and outlining image regions.

• Manual annotation. Draw polygons, rectangles, circles, lines and points for bounding boxes and segmentation datasets, with video annotation also supported.
• AI-assisted regions. Use SAM or EfficientSAM to turn selected points into polygons or masks, or create annotations from text prompts with YOLO-world and SAM3.
• Classification and review. Apply image flags for classification and dataset cleaning, with predefined labels, automatic saving and label validation.
• Dataset exports. Export VOC datasets for semantic and instance segmentation, or COCO datasets for instance segmentation.

Annotations are saved in local JSON files, and the stable annotation format can be read by your own code. The interface supports languages including English, Chinese, Japanese, Korean and Ukrainian.

LabelMe is open source under GPL-3.0 and runs on 64-bit macOS, Windows and Linux. The Python application uses Qt for its interface. A separate paid standalone app is available through labelme.io without separate Python or Qt dependencies.