CVAT – Computer Vision Annotation Tool
CVAT helps machine learning teams and AI practitioners create labeled datasets. It is designed for businesses needing a tool for various visual data types, offering open-source, SaaS, and enterprise-grade deployments.
At a glance
- Category
- Data & Analytics
- Best for
- Machine learning teams, AI practitioners, Software companies building computer vision models, Organizations requiring self-hosted data labeling
- Pricing
- CVAT Online starts at $33/month for solo users ($23/month billed annually), with Team plans starting at $66/month. An Enterprise edition starts at $12,000/year, and a free open-source Community edition is available.
- Key use cases
- Healthcare Imaging, Retail and Logistics, Autonomous Systems, Industrial Manufacturing, Agricultural Monitoring
- Integrations
- AWS S3, Google Cloud Storage, Azure Blob Storage, Hugging Face, Roboflow
- Official website
- cvat.ai/
CVAT is a visual data annotation platform used to prepare datasets for computer vision tasks. It provides tools for labeling images, videos, and 3D point clouds, supporting workflows from individual research to industrial projects.
The tool is designed for AI practitioners and software companies, including those building models for autonomous vehicles, healthcare, and manufacturing. It supports manual labeling and AI-assisted automation.
Buyers can choose between a free open-source Community edition, a SaaS-based Online version for solo users and teams, or a self-hosted Enterprise edition for organizations requiring specific data control and security.
Buyers should confirm whether they require a hosted solution or a self-managed environment and verify if specific AI model integrations, such as SAM 2 or SAM 3, align with their project requirements.
Key Features
Supports bounding boxes, polygons, points, skeletons, cuboids, and trajectories for various data types.
Integrated AI models, including SAM 2 and SAM 3, that help automate image and video segmentation.
Supports storage in AWS S3, Google Cloud Storage, and Azure Blob Storage.
The Enterprise edition includes Single Sign-On (SSO), role-based access controls (RBAC), and audit logs.
Supports manual review, ground truth jobs, and honey pots to help verify annotation accuracy.
Supports labeling of 3D point clouds and LiDAR data using 3D cuboids.
Use Cases
Labeling of medical scans for disease detection and tumor segmentation.
Creating datasets for product recognition, shelf management, and shoplifting prevention.
Labeling pedestrian and vehicle tracking data for autonomous driving and robotics.
Identification of production defects and employee activity tracking.
Annotation of crops and livestock using drone and field imagery.
Best For
Integrations
Pricing
CVAT Online starts at $33/month for solo users ($23/month billed annually), with Team plans starting at $66/month. An Enterprise edition starts at $12,000/year, and a free open-source Community edition is available.
FAQ
The Community edition is a free, self-hosted version for basic manual annotation. The Enterprise edition adds SSO, RBAC, advanced analytics, SAM integration, and dedicated technical support.
Yes, CVAT supports 3D point clouds and LiDAR data, including the use of 3D cuboids for object detection.
Yes, both the Community edition and the Enterprise edition can be self-hosted on your own infrastructure, including air-gapped environments for the Enterprise version.
CVAT integrates AI models such as SAM 2 and SAM 3 to help automate the segmentation of images and videos, which may reduce manual labeling work.
Source category: Data & Analytics
Source subcategory: Computer Vision
