NEXr Annotator

Overview

NEXr Annotator is an intelligent documentation generation platform that automatically creates comprehensive technical and business documentation directly from your source code. With the powerful BYOT (Bring Your Own Template) approach, development teams can generate customized documentation that perfectly matches their organizational standards and requirements.

Why NEXr Annotator?

Traditional Annotation Challenges

• Time-Consuming: Manual annotation is slow and tedious
• Quality Issues: Inconsistent labels across annotators
• Scalability: Difficult to manage large annotation teams
• Cost: Expensive annotation services or tools
• Workflow: Disconnected tools and processes

The NEXr Annotator Solution

• Speed: AI-assisted annotation accelerates labeling
• Quality: Built-in quality control and validation
• Collaboration: Team management and task distribution
• Cost-Effective: Flexible pricing for any team size
• Integrated: Seamless ML workflow integration

Core Capabilities

NEXr Annotator supports multiple data types and annotation tasks:

Key Features

1. Multi-Modal Annotation

Image Annotation Tools

• Bounding Boxes: Object detection and localization
• Polygon Annotation: Precise object segmentation
• Semantic Segmentation: Pixel-level classification
• Keypoint Annotation: Pose estimation and landmarks
• Instance Segmentation: Individual object masks
• Image Classification: Multi-label and single-label

Text Annotation Tools

• Named Entity Recognition (NER): Identify entities in text
• Text Classification: Categorize documents
• Sentiment Analysis: Label emotional tone
• Text Summarization: Quality evaluation
• Relation Extraction: Entity relationships
• Question Answering: QA pair creation

Video Annotation Tools

• Object Tracking: Track objects across frames
• Action Recognition: Label human actions
• Event Detection: Identify key events
• Frame-by-Frame: Detailed frame annotation
• Timeline Annotation: Temporal segmentation

Audio Annotation Tools

• Speech Transcription: Audio to text
• Speaker Diarization: Who spoke when
• Sound Classification: Identify sound types
• Emotion Recognition: Audio sentiment
• Acoustic Events: Label sound events

2. AI-Assisted Annotation

Accelerate annotation with AI:

• Auto-Labeling: Pre-label data with ML models
• Smart Suggestions: AI-powered label recommendations
• Active Learning: Focus on most valuable samples
• Model-in-the-Loop: Continuous model improvement
• Transfer Learning: Leverage pre-trained models

3. Quality Control

Ensure high-quality annotations:

• Multi-Stage Review: Annotation → Review → Approval
• Inter-Annotator Agreement: Measure consistency
• Golden Datasets: Reference standard datasets
• Quality Metrics: Track annotation quality
• Consensus Labeling: Multiple annotators per task
• Validation Rules: Custom quality checks

4. Team Collaboration

Manage annotation teams effectively:

• Role-Based Access: Annotator, Reviewer, Admin, Manager
• Task Assignment: Distribute work efficiently
• Progress Tracking: Real-time project monitoring
• Communication: In-app comments and discussions
• Performance Analytics: Annotator productivity metrics

How It Works

1. Create an Annotator Instance

Start by creating a NEXr Annotator instance from the service marketplace:

# Navigate to Service Marketplace
NEXr Cloud Dashboard → Service Marketplace → NEXr Annotator

2. Project Setup

Create Annotation Project

1. Define Project Type: Image, Text, Video, or Audio
2. Upload Data: Batch upload or connect to storage
3. Configure Labels: Define your label taxonomy
4. Set Guidelines: Create annotation instructions
5. Assign Tasks: Distribute to annotators

Label Taxonomy

{
  "project": "Object Detection - Autonomous Driving",
  "labelClasses": [
    {
      "name": "vehicle",
      "color": "#FF5733",
      "subClasses": ["car", "truck", "bus", "motorcycle"]
    },
    {
      "name": "pedestrian",
      "color": "#33FF57",
      "attributes": ["walking", "standing", "sitting"]
    },
    {
      "name": "traffic_sign",
      "color": "#3357FF",
      "subClasses": ["stop", "yield", "speed_limit"]
    }
  ]
}

3. Annotation Workflow

Use Cases

Computer Vision

Natural Language Processing

API Integration

Creating an Instance

POST /api/service-instances

{
  "instanceName": "ML Training Annotator",
  "serviceId": "nexr-annotator-service-id",
  "servicePlanId": "team-plan-id",
  "subaccountId": "your-subaccount-id",
  "globalAccountId": "your-global-account-id"
}

Creating Annotation Project

POST /api/annotation/projects

{
  "name": "Autonomous Driving Dataset",
  "type": "image_detection",
  "description": "Street scene object detection for AV training",
  "labelSchema": {
    "classes": [
      {
        "id": "vehicle",
        "name": "Vehicle",
        "color": "#FF5733",
        "subClasses": ["car", "truck", "bus", "motorcycle"]
      }
    ]
  },
  "settings": {
    "requireReview": true,
    "minAnnotatorsPerSample": 1,
    "consensusThreshold": 0.8
  }
}

Uploading Data

POST /api/annotation/projects/{projectId}/upload

{
  "dataSource": "url", // or "file", "s3", "gcs"
  "files": [
    "https://example.com/images/scene001.jpg",
    "https://example.com/images/scene002.jpg"
  ],
  "metadata": {
    "location": "San Francisco",
    "weather": "sunny",
    "time": "afternoon"
  }
}

Exporting Annotations

GET /api/annotation/projects/{projectId}/export?format=coco

# Supported formats:
# - coco (COCO JSON)
# - yolo (YOLO format)
# - pascal_voc (Pascal VOC XML)
# - csv (CSV format)
# - json (Custom JSON)

Response Format

{
  "project_id": "proj_123",
  "export_format": "coco",
  "created_at": "2024-01-15T10:00:00Z",
  "download_url": "https://storage.nexr.cloud/exports/proj_123_coco.zip",
  "statistics": {
    "total_images": 1000,
    "total_annotations": 15430,
    "label_distribution": {
      "vehicle": 8500,
      "pedestrian": 4200,
      "traffic_sign": 2730
    }
  }
}

Service Plans

Free Plan

• Pricing: Free
• Features:
  • 100 annotations/month
  • 1 annotator
  • Image and text annotation
  • Basic export formats

Starter Plan

• Pricing: $49/month
• Features:
  • 5,000 annotations/month
  • 3 annotators
  • All annotation types
  • AI-assisted labeling
  • Standard support

Team Plan

• Pricing: $199/month
• Features:
  • 25,000 annotations/month
  • 10 annotators
  • Advanced quality control
  • Custom workflows
  • Priority support
  • API access

Enterprise Plan

• Pricing: Custom
• Features:
  • Unlimited annotations
  • Unlimited annotators
  • On-premise deployment
  • Custom integrations
  • Dedicated support
  • SLA guarantees
  • Custom AI models

Enterprise Features

Security & Privacy

• Data Encryption: End-to-end encryption for sensitive data
• Access Control: Fine-grained RBAC permissions
• Audit Logs: Complete activity tracking
• Compliance: GDPR, HIPAA, SOC 2 compliant
• Data Residency: Choose data storage location

Advanced Workflows

• Custom Pipelines: Build multi-stage annotation workflows
• Automated QA: Set up automated quality checks
• Integration: Connect with MLOps platforms
• Webhooks: Real-time event notifications
• Batch Processing: Process large datasets efficiently

Analytics & Reporting

• Project Dashboards: Real-time progress tracking
• Annotator Performance: Individual productivity metrics
• Quality Reports: Inter-annotator agreement scores
• Cost Analysis: Track annotation costs per project
• Export History: Audit trail of all exports

Annotation Interface

Image Annotation Workspace

┌─────────────────────────────────────────────────────────┐
│  Project: Autonomous Driving     Progress: 45% (450/1000│
├─────────────────────────────────────────────────────────┤
│  Tools                  │  Canvas         │  Labels      │
│  ┌──────────────┐      │                 │  □ Vehicle   │
│  │ ✓ BBox       │      │   [Image]       │  □ Pedestrian│
│  │   Polygon    │      │                 │  □ Traffic   │
│  │   Keypoint   │      │                 │              │
│  │   Segmentation│      │                 │  History     │
│  └──────────────┘      │                 │  - Added bbox│
│                         │                 │  - Added bbox│
│  Attributes             │                 │              │
│  [Occluded] [Truncated]│                 │  Comments    │
│                         │                 │  💬 3 notes  │
├─────────────────────────────────────────────────────────┤
│  ← Previous    Submit    Skip    Flag    Next →         │
└─────────────────────────────────────────────────────────┘

Keyboard Shortcuts

• B: Bounding box tool
• P: Polygon tool
• K: Keypoint tool
• Delete: Remove selected annotation
• Ctrl+Z: Undo
• Ctrl+Y: Redo
• Space: Submit and next
• S: Skip current sample

Best Practices

Quality Metrics

Inter-Annotator Agreement

Measure consistency between annotators:

• Cohen's Kappa: Agreement between two annotators
• Fleiss' Kappa: Agreement among multiple annotators
• IoU (Intersection over Union): Bounding box overlap
• Dice Coefficient: Segmentation mask similarity

Performance Tracking

Monitor annotation team performance:

{
  "annotator": "user_123",
  "period": "2024-01-01 to 2024-01-31",
  "metrics": {
    "annotationsCompleted": 2450,
    "averageTimePerSample": "45 seconds",
    "accuracy": 0.94,
    "reviewRejectionRate": 0.06,
    "productivity": "High"
  }
}

Integrations

ML Frameworks

• TensorFlow: Direct dataset export
• PyTorch: DataLoader compatible formats
• Keras: Generator-ready exports
• Hugging Face: Dataset Hub integration

Cloud Storage

• AWS S3: Direct upload/download
• Google Cloud Storage: Native integration
• Azure Blob Storage: Seamless connection
• MinIO: Self-hosted object storage

MLOps Platforms

• MLflow: Experiment tracking integration
• Weights & Biases: Dataset versioning
• DVC: Data version control
• Label Studio: Import/export compatibility

Support & Resources

Launch & Access

Web Interface

Access through NEXr Cloud dashboard:

NEXr Cloud → Service Instances → NEXr Annotator → Launch

Programmatic Access

Use service keys for API integration:

const response = await fetch('https://nexr-annotator.nexr.cloud/api/v1/projects', {
  headers: {
    'Authorization': `Bearer ${serviceKey}`,
    'Content-Type': 'application/json'
  }
});

Authentication Flow

1. User clicks "Launch" in NEXr Cloud dashboard
2. System generates time-limited launch token (5 minutes)
3. User is redirected to Annotator with token
4. Annotator validates token and creates session
5. User accesses project with proper permissions

Getting Started

Ready to build high-quality ML datasets?