What is Labelme used for?

Labelme is a desktop image annotation tool that runs SAM2, SAM3 and YOLO-World models locally to draw polygons and boxes and export YOLO and Pascal VOC data. It is a good fit for preparing computer vision training datasets and annotating sensitive images that cannot leave a machine.

How much does Labelme cost?

Paid plans for Labelme start at $49 per seat as a one-time payment, and there is no free plan. A free trial is available.

Is Labelme open source?

No. Labelme is proprietary (closed-source) software. In the AI Infrastructure category, open-source options include Label Studio, whisper.cpp and llamafile.

Can I self-host Labelme?

Yes. Although Labelme is closed source, it can be self-hosted on your own servers.

What are some alternatives to Labelme?

Labelme competes with Labelbox, Roboflow and Encord.

Labelme: features, pricing and alternatives

About Labelme

Labelme is an AI-assisted image annotation app for macOS, Windows and Linux that keeps all work on your computer. Images, annotations and exports are never uploaded, and the AI models run locally: SAM2, SAM3 and YOLO-World are downloaded once and then work offline, with no API key or third-party AI account. It comes from the creator of the open-source Labelme project.

The workflow is to open a folder of images, outline objects with a few clicks, check the labels and export training-ready files in YOLO and Pascal VOC formats. Annotation can be prompted with a point (click to include, shift-click to exclude), a rough box that returns an oriented box for every similar object, or a text phrase that annotates all matching objects. Every result remains editable point by point, and polygons, bounding boxes and segmentation are supported.

Labelme is sold as a one-time purchase, with Starter and Pro tiers and a Pro dataset toolkit for checking and QA before export. It is available as a free 3-day trial. The site is available in English, Japanese and Chinese and lists agriculture, medical, industrial and traffic images among typical uses.

Key features

Local SAM2, SAM3 and YOLO-World models
Point, box and text prompts for annotation
Polygon, bounding box and segmentation labels
Editable results point by point
Export to YOLO and Pascal VOC formats
Offline work after one-time model download
Dataset checks in the Pro toolkit

Good fit for

Preparing computer vision training datasets
Annotating sensitive images that cannot leave a machine

Tags: image-annotation; computer-vision; sam2; yolo; offline; desktop; dataset; privacy

Labelme: questions and answers

What is Labelme used for?: Labelme is a desktop image annotation tool that runs SAM2, SAM3 and YOLO-World models locally to draw polygons and boxes and export YOLO and Pascal VOC data. It is a good fit for preparing computer vision training datasets and annotating sensitive images that cannot leave a machine.
How much does Labelme cost?: Paid plans for Labelme start at $49 per seat as a one-time payment, and there is no free plan. A free trial is available.
Is Labelme open source?: No. Labelme is proprietary (closed-source) software. In the AI Infrastructure category, open-source options include Label Studio, whisper.cpp and llamafile.
Can I self-host Labelme?: Yes. Although Labelme is closed source, it can be self-hosted on your own servers.
What are some alternatives to Labelme?: Labelme competes with Labelbox, Roboflow and Encord.