Annotate better with CVAT, the data engine trusted by machine learning teams at any scale, for data of any scale. Build datasets faster with AI-powered tools, collaborate seamlessly, and deploy anywhere.
Follow these steps to start annotating your first dataset
1
Sign up for CVAT
Create a free account on CVAT Cloud or install CVAT on your own infrastructure. Cloud accounts include 10 free tasks and 500MB of storage to get you started.
2
Create your first project
Set up a project with custom labels and attributes for your annotation task. Projects help you organize related tasks and maintain consistent labeling schemes.
# Using the CLIcvat-cli --auth user:password project create \ --labels '[{"name": "car"}, {"name": "person"}]' \ "My First Project"
3
Upload and annotate data
Upload images or videos to create tasks within your project. Use the intuitive annotation editor to draw bounding boxes, polygons, polylines, points, or 3D cuboids.
CVAT supports 30+ annotation formats including COCO, YOLO, Pascal VOC, Cityscapes, and more.
4
Leverage AI-powered auto-annotation
Speed up annotation by up to 10x using built-in models like Segment Anything (SAM), YOLO, or Mask R-CNN. Or integrate your own models via serverless functions.
from cvat_sdk import Clientclient = Client(url="https://app.cvat.ai")client.login(("username", "password"))task = client.tasks.retrieve(task_id)# Auto-annotate using built-in models
Explore by Topic
Everything you need to build world-class datasets
Annotation Tools
Master the annotation editor with support for images, videos, and 3D point clouds
Auto-Annotation
Leverage AI models to speed up annotation with Segment Anything, YOLO, and more
Dataset Formats
Import and export in 30+ formats including COCO, YOLO, Pascal VOC, and Cityscapes
REST API
Integrate CVAT into your ML pipeline with our comprehensive REST API
Python SDK
Automate workflows with the powerful Python SDK for programmatic access
Self-Hosted Setup
Deploy CVAT on your infrastructure with Docker, Kubernetes, or AWS
