Windows downloads are now digitally signedv7.4.1
For Healthcare & Research

Annotate medical images.
Nothing leaves your machine.

Labelme runs entirely on your machine. Patient images never leave your network — no cloud uploads, no third-party access, no compliance headaches.

Get Labelme Pro →

Install and start annotating in minutes.

HIPAA-friendly architecture
Zero cloud upload
No BAA required
Offline after model download
5.0·940+ 用户·GitHub 16K+ 星

Cloud annotation tools create compliance risks

  • Uploading patient images to external servers may violate HIPAA, GDPR, or institutional data governance policies
  • BAA agreements add procurement overhead and delay project starts
  • Air-gapped hospital networks and secure research environments block internet-dependent tools entirely

Labelme is installed locally and processes everything on-device. No data leaves your machine.

Built for every medical imaging workflow

Pathology

Annotate tissue samples for cancer detection and cell classification models.

H&E slidesWhole-slide images (PNG/TIFF)Cell segmentationTumor regions

Radiology & ophthalmology

Label retinal scans, X-ray screenshots, and exported imaging frames for diagnostic AI.

Retinal fundus photosX-ray PNG exportsLesion detectionOrgan segmentation

Endoscopy & surgical video

Extract and annotate frames from endoscopic or surgical footage.

Polyp detectionInstrument trackingAnatomy labelingVideo frames

Dermatology

Mark lesions, wound boundaries, and skin conditions for classification datasets.

Lesion boundaryMelanoma screeningWound progression

Built for privacy-sensitive environments

Offline annotation

Runs on Windows, macOS, and Linux with no internet connection once the AI models are downloaded.

No cloud storage

Your images are never uploaded anywhere. Annotations are saved as JSON files on your machine.

Works in air-gapped networks

Download the AI models once on a connected machine, then run anywhere, including secure hospital or lab environments.

AI-assisted annotation

One-click segmentation with SAM2 (Segment Anything). Runs locally — no external API calls.

Standard export formats

Export to YOLO, Pascal VOC, or custom formats compatible with PyTorch, TensorFlow, and MONAI.

One-time purchase

No subscription. One license covers one person on all their own machines. Pay once, use indefinitely.

FAQ

Is Labelme HIPAA compliant?

Labelme is an annotation tool, not a covered entity or business associate. Because it processes all data locally and never transmits patient images externally, it removes the compliance risk that comes with cloud-based tools. Your institution's legal team can verify: no data leaves your machine.

Can I use it in an air-gapped or offline hospital network?

Yes, after a one-time download per AI model — download the models on a connected machine first, or ask support@labelme.io for the offline pack. Once the models are on disk, Labelme needs no internet connection. You can run it on a local machine, a secure workstation, or even a USB drive.

How do I annotate DICOM files?

Convert your DICOM files to PNG using a tool like pydicom or 3D Slicer, then open them in Labelme. The conversion takes seconds and Labelme works with all standard image formats (PNG, JPEG, BMP, TIFF).

Can our research lab get a volume license or invoice?

Yes. We support institutional purchases, purchase orders, and invoices. Email support@labelme.io with your institution name, number of seats, and any procurement requirements.

What formats can I export annotations to?

Labelme saves to Labelme JSON (default). The Pro toolkit adds YOLO and Pascal VOC exports with batch processing. These formats work with PyTorch, TensorFlow, Detectron2, and MONAI.

Ready to annotate without compliance headaches?

One-time purchase. No subscription. Works offline once the AI models are downloaded.

Get Labelme Pro →
14-day money-back guarantee
·
Secure payment via Paddle

Questions? Email support@labelme.io