Creating Projects

Learn how to create and configure CVAT projects with labels, attributes, and settings

Projects in CVAT help you organize related annotation tasks with shared labels, settings, and quality control configurations. A project serves as a container for multiple tasks that share the same annotation schema.

Overview

A project contains:

• Shared label schema: Labels and attributes used across all tasks
• Quality settings: Validation and quality control configuration
• Tasks: Individual annotation jobs with uploaded data
• Organization: Optional workspace for team collaboration

Creating a Project

Using the Web UI

1. Navigate to the Projects page
2. Click Create new project
3. Enter project details:
   • Name: Descriptive project name
   • Labels: Define your annotation schema (see Labels & Attributes)
   • Bug tracker: Optional issue tracking URL
4. Configure advanced settings (optional):
   • Source storage: Cloud storage for importing datasets
   • Target storage: Cloud storage for exporting annotations
5. Click Submit to create the project

Using the REST API

Create a project with the POST /api/projects endpoint:

curl -X POST "https://app.cvat.ai/api/projects" \
  -H "Authorization: Token <your-token>" \
  -H "Content-Type: application/json" \
  -d '{
    "name": "Vehicle Detection Project",
    "labels": [
      {
        "name": "car",
        "color": "#ff0000",
        "type": "rectangle",
        "attributes": [
          {
            "name": "occluded",
            "input_type": "checkbox",
            "mutable": true,
            "values": ["false"]
          }
        ]
      },
      {
        "name": "pedestrian",
        "color": "#00ff00",
        "type": "rectangle"
      }
    ]
  }'

Response (201 Created):

{
  "id": 1,
  "name": "Vehicle Detection Project",
  "owner": {
    "id": 1,
    "username": "admin",
    "first_name": "John",
    "last_name": "Doe"
  },
  "assignee": null,
  "bug_tracker": "",
  "created_date": "2026-03-04T10:00:00.000Z",
  "updated_date": "2026-03-04T10:00:00.000Z",
  "status": "annotation",
  "dimension": null,
  "organization_id": null,
  "tasks": {
    "count": 0,
    "url": "https://app.cvat.ai/api/tasks?project_id=1"
  },
  "labels": {
    "url": "https://app.cvat.ai/api/labels?project_id=1"
  }
}

Using the Python SDK

Install the SDK:

pip install cvat-sdk

Create a project:

from cvat_sdk import Client, models

# Initialize client
client = Client(url="https://app.cvat.ai")
client.login(("username", "password"))

# Define labels
labels = [
    models.PatchedLabelRequest(
        name="car",
        color="#ff0000",
        type="rectangle",
        attributes=[
            models.AttributeRequest(
                name="vehicle_type",
                input_type="select",
                mutable=False,
                values=["sedan", "suv", "truck"],
                default_value="sedan"
            )
        ]
    ),
    models.PatchedLabelRequest(
        name="traffic_light",
        color="#ffff00",
        type="rectangle"
    )
]

# Create project
project = client.projects.create(
    spec=models.ProjectWriteRequest(
        name="Traffic Scene Analysis",
        labels=labels
    )
)

print(f"Created project ID: {project.id}")
print(f"Project name: {project.name}")

Using the CLI

Install the CLI:

pip install cvat-cli

Create a project:

# Basic project creation
cvat-cli --auth user:password project create \
  --labels '[{"name": "person"}, {"name": "vehicle"}]' \
  "Pedestrian Detection"

# Create project with label attributes
cvat-cli --auth user:password project create \
  --labels '[
    {
      "name": "car",
      "color": "#ff0000",
      "attributes": [
        {
          "name": "color",
          "input_type": "select",
          "values": ["red", "blue", "white", "black"]
        }
      ]
    }
  ]' \
  "Vehicle Analysis"

# Create project from existing dataset
cvat-cli --auth user:password project create \
  --labels labels.json \
  --dataset_path ./coco_dataset.zip \
  --dataset_format "COCO 1.0" \
  "COCO Import Project"

Project Configuration

Organization Assignment

Projects can belong to an organization for team collaboration:

# Create project in an organization
project = client.projects.create(
    spec=models.ProjectWriteRequest(
        name="Team Project",
        labels=labels,
        organization_id=5  # Organization ID
    )
)

Only organization members can access projects within that organization. Projects without an organization are personal projects.

Storage Configuration

Configure cloud storage for importing and exporting data:

from cvat_sdk import models

project = client.projects.create(
    spec=models.ProjectWriteRequest(
        name="Cloud Storage Project",
        labels=labels,
        source_storage=models.StorageWrite(
            location="cloud_storage",
            cloud_storage_id=10
        ),
        target_storage=models.StorageWrite(
            location="cloud_storage",
            cloud_storage_id=11
        )
    )
)

Storage options:

• local: CVAT server storage (default)
• cloud_storage: AWS S3, Azure Blob, or Google Cloud Storage

Bug Tracker Integration

Link an external issue tracker:

curl -X POST "https://app.cvat.ai/api/projects" \
  -H "Authorization: Token <your-token>" \
  -H "Content-Type: application/json" \
  -d '{
    "name": "Quality Control Project",
    "bug_tracker": "https://github.com/org/repo/issues",
    "labels": [...]
  }'

Project Properties

Key project attributes returned by the API:

Field	Type	Description
id	integer	Unique project identifier
name	string	Project name
owner	object	User who created the project
assignee	object	User assigned to the project
status	string	Project status: annotation, validation, completed
created_date	datetime	Creation timestamp
updated_date	datetime	Last modification timestamp
dimension	string	Task dimension: 2d or 3d
organization_id	integer	Organization ID (null for personal projects)
tasks	object	Summary of project tasks
labels	object	Project label schema

Best Practices

Design your label schema carefully

• Define all labels upfront before creating tasks
• Use consistent naming conventions
• Add attributes for important object properties
• Consider label hierarchies for complex scenarios
• Review Labels & Attributes guide

Organize with meaningful names

• Use descriptive project names: “Traffic Scene Detection Q1 2026”
• Include version numbers for iterative datasets
• Add date ranges or data source information

Configure quality settings early

• Set up validation parameters at project creation
• Define quality thresholds before annotation begins
• Enable consensus annotation for critical tasks
• See Quality Control for details

Use organizations for team work

• Create projects within organizations for collaboration
• Assign roles and permissions appropriately
• Configure shared cloud storage for large datasets

Next Steps

Managing Tasks

Create and manage annotation tasks within your project

Labels & Attributes

Configure labels, attributes, and annotation schemas

Quality Control

Set up validation workflows and quality metrics

API Reference

View complete API documentation