opencv/cvat
Cvat
CVAT 是一个开源计算机视觉标注工具和可视化数据集管理平台。它提供了一个自托管界面，用于标注图像、视频和 3D 数据，以创建视觉 AI 模型的数据集。
该平台具有 AI 辅助数据标注功能，可自动创建掩码和边界框，并利用插件系统连接外部机器学习模型。它包括一个基于共识的质量保证系统，通过比较独立标注来验证标签准确性。
该系统涵盖协作团队管理、通过任务分解进行项目组织以及远程云存储集成。它还提供用于程序化工作流控制以及以行业标准格式导入和导出数据的 REST API。
Features
- Computer Vision Tools - Provides an interactive software interface specifically designed for labeling and preparing visual datasets for computer vision models.
- Visual Annotation Tools - Provides specialized drawing tools for labeling images, videos, and 3D objects to create computer vision datasets.
- AI-Assisted Labeling - Uses artificial intelligence to automate the labeling process and reduce manual work for computer vision tasks.
- Computer Vision Annotation - Creates high-quality annotated datasets for images, videos, and 3D objects to train and evaluate vision AI models.
- Dataset Management Tools - Implements utilities for organizing, annotating, and converting visual datasets to support machine learning training pipelines.
- Model-Assisted Labelers - Utilizes machine learning models to automatically generate initial bounding boxes and masks for visual data.
- Automated Annotations - Provides AI-powered automation to suggest or create visual annotations, significantly reducing manual labeling effort.
- Self-Hosted Labeling Platforms - Offers a self-hosted, open-source environment for managing annotation workflows and collaborative labeling tasks.
- Consensus-Based Label Verification - Implements a consensus-based system to verify label accuracy by comparing independent annotations from different users.
- Visual Label Verification - Provides mechanisms to review and verify label accuracy through consensus checks and ground truth comparisons.
- AI Model Integrations - Features a plug-in system to connect external machine learning models to the labeling interface for automated suggestions.
- Dataset Partitioning - Organizes massive datasets into smaller manageable units assigned to specific users for scalable labeling and review.
- Annotation Project Management - Provides a centralized environment for organizing visual data into tasks and managing collaborative labeling workflows.
- Cloud Storage Integrations - Provides a connection bridge to remote storage providers for managing large datasets without manual upload processes.
- Data Import and Export - Supports transferring data between the system and industry-standard file formats to ensure cross-environment compatibility.
- Remote Object Storage Integrations - Connects to cloud buckets and external file systems to stream large datasets without requiring full local storage.
- Team Management - Provides tools for organizing team members, assigning roles, and tracking collective progress analytics.
- Workflow Automation APIs - Provides programmatic interfaces for managing task creation, data uploads, and exports via automated scripts.
- Background Job Processing - Handles computationally expensive data import and export tasks in the background to maintain a responsive user interface.
- Role-Based Access Control - Manages data visibility and editing permissions through predefined user roles to secure collaborative workflows.
- Client-Server Architectures - Utilizes a client-server architecture that separates the annotation interface from the data processing backend via a REST API.
- Collaborative Annotation - Manages shared data access through role-based permissions and integrated discussion threads for collaborative labeling.
- Annotation Quality Verifications - Ships a consensus-based quality assurance system that verifies label accuracy by comparing independent annotations.
- Annotation - Efficient computer vision annotation tool.
- Annotation and Data Tools - Computer Vision Annotation Tool.
- Computer Vision Libraries - Powerful tool for computer vision data annotation.
- 数据标注 - 高效的计算机视觉标注工具。
- Image Annotation - High-performance platform for computer vision annotation tasks.
