Supervisely

Supervisely is an all-in-one computer vision platform: curate data, annotate it with AI help, review the labels, then train, evaluate and deploy models on the same project without exporting between tools. It runs as a cloud service or fully self-hosted, including offline installs, which makes it a common choice for enterprises with data that cannot leave their network.

Labeling

• Editors for images, video (tracking, multi-camera and GPS telemetry), 3D LiDAR point clouds with sensor fusion and AI cuboids, DICOM and NIfTI medical volumes, multispectral and microscopy images, meshes and audio.
• AI assistants pre-label whole datasets, label from a text prompt or a single example, and learn live from corrections (active learning); the Smart Tool segments an object from a few clicks.
• Labeling jobs, queues, review, consensus, annotator exams and per-person speed and quality dashboards for managing labeling teams.

Data and models

• Import COCO, YOLO and other formats, or link files in S3, GCS or Azure, then search a project by meaning with AI Search and manage collections and versions.
• Fine-tune detection and segmentation models on your labels using your own GPU machines (no paid training credits), with live charts and experiment history.
• Every training run ends in an evaluation report — mAP, precision, recall, errors by class and by image — and deployed models pre-label the next batch.
• An ecosystem of open-source apps and a Python SDK extend the platform.

Pricing

Community is free for individuals, researchers and small teams: 2 members, 5 GB of storage and 10,000 files, with 500 Smart Tool requests a day. Pro starts at €199 a month with 3 members, 50 GB and unlimited retention, plus cheaper labeling-only seats. Enterprise is custom, self-hosted or cloud, with SSO, custom roles and an SLA.