LabelMe

LabelMe is a graphical image annotation tool for building labeled image datasets from images and video. It helps with polygon, rectangle, circle, line, and point annotations, plus classification flags and cleaning. The free pip-installed Python and Qt tool works 100% offline, with a paid standalone build that needs no Python or Qt.

It exports VOC-format data for semantic segmentation and instance segmentation, and COCO-format data for instance segmentation. AI-assisted annotation includes point-to-polygon and point-to-mask labeling with SAM and EfficientSAM, plus text-to-annotation with YOLO-World and SAM3. GUI options include predefined labels and flags, auto-saving, and label validation.

LabelMe is written in Python with a Qt interface and is available in 20 languages. A paid standalone build on labelme.io installs without Python or Qt, and because everything runs locally and offline, it suits air-gapped or privacy-sensitive workflows where data must stay on the machine.