Appearance

Image Labeling ​

What is Labeling? ​

Labeling (or annotation) is the process of marking images with information that allows AI models to learn to recognize patterns. It is the fundamental basis for training effective machine learning models.

Access to Labeling ​

From a Dataset ​

  1. Open dataset from the main view
  2. "Labeling" tab at the top
  3. "Labeling" button with image icon

Editor Interface ​

The labeling tool opens in full screen with:

  • Main canvas: Where the image is displayed
  • Toolbar: Annotation tools
  • Navigation panel: List of images in the dataset
  • Information panel: Current image details

Labeling Tools ​

🖼️ Basic Tools ​

Visualization ​

  • 🔍 Zoom: Zoom in and out of the image
  • 👋 Pan: Move the view around the image
  • 🔄 Rotation: Rotate the image for better visualization
  • 🎨 Filters: Adjust brightness, contrast for better visibility
  • ⏮️ Previous: Go to previous image
  • ⏭️ Next: Go to next image
  • 📋 List: View all images in the dataset
  • 🔢 Go to image: Jump to a specific image

🎯 Tools by Dataset Type ​

For Classification ​

  • 🏷️ Class selector: Dropdown to choose category
  • 💾 Auto-save: Classification is saved when selected
  • 📊 Progress: Counter of classified images

For Object Detection ​

  • 📦 Bounding Box: Create bounding boxes
  • ✏️ Edit boxes: Resize and move existing boxes
  • 🗑️ Delete: Remove incorrect annotations
  • 🏷️ Label boxes: Assign class to each box

For Segmentation ​

  • ✏️ Brush: Paint areas of interest
  • 🧽 Eraser: Remove parts of the selection
  • 🪄 Magic tool: Automatic selection by color
  • 📐 Polygon: Create precise shapes with points

For Anomaly Detection ​

  • ✅ Normal: Mark image as normal (no defects)
  • ❌ Anomalous: Mark image as anomalous
  • 📝 Notes: Add comments about the defect

Advanced Features ​

🤖 AI Assistance ​

Segment Anything Model (SAM) ​

  • Click selection: Click on object for automatic segmentation
  • Refinement: Automatically adjust edges
  • Smart suggestions: Segmentation proposals

Vision-Language Models (VLM) ​

  • Automatic description: Generate AI labels
  • Guided detection: Find objects by text description
  • Smart validation: Verify annotation consistency

⚡ Productivity Tools ​

Keyboard Shortcuts ​

  • Space: Zoom fit
  • +/-: Zoom in/out
  • Arrows: Navigate between images
  • Esc: Cancel current tool
  • Ctrl+Z: Undo last action

Batch Annotation ​

  • Multiple selection: Apply label to multiple images
  • Quick filters: View only unlabeled images
  • Copy annotations: Between similar images

Image Editor ​

🛠️ Editing Tools ​

Basic Adjustments ​

  • ☀️ Brightness: Lighten or darken
  • 🌓 Contrast: Improve definition
  • 🌈 Saturation: Color intensity
  • ⚖️ Gamma: Exposure correction

Filters ​

  • 🔳 Grayscale: For better analysis
  • 🔍 Edge enhancement: Highlight contours
  • 📊 Histogram: Color distribution analysis
  • 🎭 Custom filters: According to analysis type

💾 Change Management ​

History ​

  • 📚 Change stack: Record of all modifications
  • ↩️ Undo: Revert changes one by one
  • ↪️ Redo: Restore undone changes
  • 💾 Save points: Create intermediate versions

Saving ​

  • 🔄 Automatic: Annotations are saved automatically
  • ☁️ Synchronization: With database in real time
  • 📤 Export: Download annotations in different formats

Validation and Quality ​

✅ Annotation Validation ​

Automatic Review ​

  • Inconsistency detection: Contradictory annotations
  • Boundary verification: Boxes outside the image
  • Class balance: Warnings about imbalance
  • Completeness: Unannotated images

Quality Metrics ​

  • 📊 Percentage completed: Labeling progress
  • ⚖️ Class distribution: Dataset balance
  • 🎯 Precision: Consistency in annotations
  • ⏱️ Average time: Labeling efficiency

🔍 Review Tools ​

Summary View ​

  • 📋 Annotation list: All dataset labels
  • 🔍 Search: Find specific annotations
  • 📊 Statistics: Metrics by class and type
  • 📈 Charts: Distribution visualization

Best Practices ​

🎯 Consistency ​

  • Clear criteria: Define labeling rules before starting
  • Documentation: Maintain updated annotation guides
  • Cross reviews: Validate work between different annotators
  • Regular calibration: Check consistency periodically

⚡ Efficiency ​

  • Keyboard shortcuts: Learn and use shortcuts frequently
  • Logical order: Label in coherent sequences
  • Appropriate tools: Use the right tool for each type
  • Regular breaks: Avoid visual fatigue

🎨 Precision ​

  • Appropriate zoom: Adequate detail level for each task
  • Precise edges: Exact annotations at boundaries
  • Edge cases: Document ambiguous situations
  • Final review: Verify before marking as complete

Common Problems ​

🐛 Technical Issues ​

  • Canvas not responding: Refresh browser or clear cache
  • Annotations not saving: Check internet connection
  • Tools disabled: Check user permissions
  • Images not loading: Verify file format and size

🎯 Quality Issues ​

  • Inconsistent annotations: Review labeling criteria
  • Class imbalance: Ensure equitable representation
  • Ambiguous cases: Establish rules for edge situations
  • Image quality: Improve source lighting or resolution