cvat-ai/cvat

Features

• Computer Vision Platforms - Serves as a comprehensive web-based platform for annotating images, videos, and 3D point clouds for machine learning.
• Annotation Tools - Provides comprehensive tools for manually creating precise polygonal shapes for image segmentation.
• Automated Annotation Tools - Provides automated object detection and segmentation using serverless machine learning models to accelerate the labeling process.
• Data Annotation Workflows - Provides a comprehensive environment for managing the entire lifecycle of data annotation tasks, including team collaboration and quality control.
• Computer Vision Tools - Offers an interactive web-based interface for labeling and preparing visual datasets for machine learning model training.
• Model-Assisted Labelers - Applies pre-trained machine learning models to generate initial annotations or suggest labels, reducing manual effort.
• Dataset Management - Enables the import, organization, and export of annotated datasets in standard formats for machine learning pipelines.
• Dataset Management Tools - Exports annotated datasets into structured formats including geometric shapes, attributes, and tracking identifiers for machine learning training.
• Annotation Project Management - Configures data sources, label schemas, and processing parameters to define new labeling projects.
• Containerized Annotation Servers - Deploys as a containerized application to orchestrate annotation tasks, model-assisted labeling, and data storage.
• RESTful API Interfaces - Provides a standardized programmatic layer for managing projects, users, and annotation data through structured HTTP requests.
• Annotation Consensus Mergers - Ensures high-quality ground truth by merging diverse labeling opinions into a consensus dataset.
• Policy-Based Access Control - Enforces granular security permissions and authorization rules using a centralized policy engine.
• Role-Based Access Control - Enforces granular security policies by validating user permissions and organizational memberships against a centralized authorization engine.
• Self-Hosted Labeling Platforms - Functions as a self-hosted, open-source environment for managing annotation workflows and team collaboration.
• Testing & Quality Assurance - Ensures high-quality ground truth data through consensus mechanisms, conflict detection, and detailed quality assurance metrics.
• Annotation Grouping Tools - CVAT associates multiple individual shapes into a single logical unit to simplify dataset organization and enable collective metadata management.
• Annotation Interpolation Tools - Provides automated point movement interpolation to track objects across video frames.
• 3D - Enables copying, cutting, and pasting cuboids within three-dimensional space to replicate or reposition labels.
• 3D Object Labeling - Supports creating three-dimensional cuboid annotations by defining planar surfaces and depth.
• Bulk Annotation Managers - Accelerates labeling workflows through bulk operations and label propagation across frames.
• Skeleton Annotation Tools - Supports drawing and tracking multi-point skeleton structures for human pose estimation and similar tasks.
• Geometry Modification Tools - Enables precise refinement of annotation shapes by combining or splitting entities.
• Object Mask Generators - Paints pixel-level masks over image regions to create precise boundaries for object segmentation tasks.
• Annotation Integration Pipelines - Maps remote annotation data into standard machine learning structures to create seamless training pipelines for deep learning models.
• Video Object Tracking - Implements automated object tracking to propagate segmentation masks and shapes across sequential video frames.
• Label Definition Managers - Standardizes annotation categories by managing the creation, retrieval, and updates of label definitions.
• Backup and Recovery - Maintains data integrity through automated backup and system upgrade processes.
• Consensus Replica Generators - Captures diverse annotator perspectives by generating multiple instances for the same dataset.
• Remote Storage Mounts - Connects remote object storage buckets directly to the workspace to enable seamless access to large-scale datasets.
• Manual Annotation Management - Provides interactive tools to draw shapes and apply tags to images or video frames to define ground truth data.
• Media Pipeline Integrations - Connects to remote storage to import raw media and export finished annotations directly into machine learning pipelines.
• Enterprise Deployment Solutions - Orchestrates containerized services to support scalable, enterprise-grade production deployments across various infrastructure providers.
• Kubernetes Deployments - Provides native support for orchestrating platform components across Kubernetes clusters for high availability and scalability.
• Model Inference Functions - Executes external machine learning models as remote, event-driven tasks to automate data labeling and object tracking.
• Automated HTTPS Management - Automates the acquisition and renewal of TLS certificates for secure web access.
• SAML SSO Integrations - Integrates external identity providers using OIDC or SAML protocols to manage user login through centralized authentication services.
• Annotation Attributes - Attaches descriptive metadata to annotation labels to provide additional context for machine learning datasets.
• Microservices Architectures - Orchestrates platform components as isolated, scalable services within container environments to ensure consistent deployment.
• User Account Management - Provides tools to manage user profiles and account access within the annotation platform.
• Annotation Analytics Controllers - Tracks labeling accuracy and performance through consensus metrics and analytics.
• Validation Dataset Managers - Enables automated quality estimation through the management of validation and ground truth datasets.
• Annotation Layering Controls - CVAT organizes overlapping objects into distinct Z-axis layers and provides controls to navigate, add, or reorder objects across these depth levels.
• Annotation Visibility Controls - CVAT allows users to lock specific objects to prevent accidental modification and adjust transparency to simplify labeling in frames with high object density.
• Skeleton Annotation Tools - Supports automated skeleton point detection and mapping for human pose or object structure annotation.
• Data Annotation and Synthesis - Web-based annotation tool for images and videos.
• Organization-Based Access Management - Manages workspace access and resource sharing based on organizational membership.
• Quality Report Exporters - Provides detailed quality reports for automated processing and external review.
• Team Collaboration Management - Organizes users and datasets into distinct workspaces to manage access control and team collaboration.
• Schema Definition - Defines structured labels and properties to standardize data for machine learning models.
• Relational State Management - Persists project configurations, user metadata, and annotation records in a relational database to ensure data integrity.
• Spatial Data Labeling - Labels volumetric objects within three-dimensional point clouds or depth-aware data to support spatial perception tasks.
• Project Lifecycle Managers - Manages the lifecycle of annotation tasks including backup, migration, and deletion.
• Annotation Progress Trackers - Tracks the progress and quality of annotation work through defined stages and status updates.
• Cloud Storage Integrations - Links external cloud storage providers to the system to list, preview, and access remote data assets for annotation workflows.
• Video Frame Navigators - Provides precise controls for stepping through, seeking, and inspecting individual frames within video sequences.
• Access Token Management - Creates, lists, updates, and revokes API tokens to provide programmatic access to platform resources.
• Session Authentication - Validates user credentials to establish a session and issues a token for subsequent API requests.
• Social Login Providers - Integrates third-party identity services like Google and GitHub to allow users to sign in using existing external accounts.
• Background Processing - Manages long-running operations like data imports and model inference through distributed job queues to maintain system responsiveness.
• Webhook Event Notifications - Sends automated event updates to external endpoints by registering subscription URLs for specific system actions.
• Segmentation Mask Handlers - Supports creating pixel-level segmentation masks for disjoint objects using brush-based painting and conversion tools.
• Interface Labeling - Provides interactive tools for drawing and modifying geometric shapes to label objects in datasets.
• Custom AI Models - Allows deployment of custom machine learning models as serverless functions for inference within the annotation workspace.
• Annotation Guidelines - Establishes structured guidelines and requirements to ensure consistency across annotation teams.
• Annotation Conversion Tools - Transforms labels between geometric formats to standardize dataset representations.
• Data Archiving Systems - Enables the creation of compressed snapshots for long-term project preservation and disaster recovery.
• User Management - Provides an administrative interface to reset user passwords and restore account access when automated recovery methods are unavailable.
• Webhook Security - Validates the authenticity and integrity of outgoing webhook payloads using cryptographic signatures.
• Programmatic Interfaces - Executes administrative actions and manages project data through a structured interface to perform complex workflows programmatically.
• Workflow Stage Assignments - Designates specific workflows like validation or acceptance to team members to define responsibilities.
• Annotation Collaboration - Provides collaborative tools for attaching feedback and notes to specific annotation tasks to facilitate team communication.
• Geometric Snapping Tools - Automatically snaps polygon points to the vertices of nearby shapes to ensure precise alignment.
• Mask Overlap Handlers - Automatically deletes underlying pixels when adding new masks to ensure clean boundaries between adjacent objects.
• Annotation Workflow Accelerators - Reduces manual effort through customizable keyboard shortcuts and interface optimizations.
• Asset Type Definitions - Specifies source data formats to ensure the annotation interface correctly handles video, images, or audio.
• Data Categorization - Allows users to assign descriptive tags to entire images or video frames to classify content without spatial coordinates.
• Linear Feature Labeling - Draws paths using sequences of points to label linear structures like road markings.
• Cloud Storage Authentication - Integrates remote cloud storage providers by managing authentication and resource paths for direct data access.
• Outsourcing Workflows - Distributes labeling jobs to decentralized networks and manages compensation through protocol-based systems.
• Playback Preloaders - Preloads data chunks in the background to ensure smooth navigation and playback of large video sequences.
• Visual Enhancement Utilities - Modifies brightness, contrast, saturation, and gamma levels to improve visibility during the annotation process.
• API Client Generators - Fetches machine-readable interface specifications to facilitate integration with external tools and automated client code generation.
• Audit Logging - Records and monitors system events and user actions to maintain compliance and security within the annotation environment.
• Background Job Queues - Manages asynchronous job queues for data processing, auto-annotation, and webhook delivery to ensure system responsiveness.
• Health Monitoring - Monitors container status and disk usage to ensure platform operational health.

Star history