Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/jusolo/labelWise/llms.txt

Use this file to discover all available pages before exploring further.

​
Getting Started Right

​
Organize Before You Annotate

Before loading images into labelWise, organize them in folders by similarity (same scene, same object types, similar lighting). This helps you work in batches and reuse annotations.
Preparation steps:
  1. Group similar images together
  2. Ensure consistent naming conventions
  3. Remove any corrupted or duplicate images
  4. Verify all images are in supported formats (JPEG, PNG, WebP, etc.)

​
Efficient Annotation Techniques

​
Use Copy and Paste for Similar Objects

The most powerful workflow optimization in labelWise is the copy/paste feature:
Template workflow: Create a perfect annotation once, then copy and paste it across similar images. Adjust the position and size as needed.
Example workflow:
  1. Annotate a car in the first image with precise boundaries
  2. Press Ctrl/Cmd + C to copy
  3. Navigate to the next image with a similar car
  4. Press Ctrl/Cmd + V to paste
  5. Drag to reposition and resize if needed
  6. Repeat for all similar images

​
Multi-Select for Batch Operations

Use Ctrl/Cmd + Click to select multiple annotations at once. This is perfect for:
  • Copying multiple objects together
  • Reviewing multiple annotations before moving on
  • Deleting multiple incorrect annotations
Batch workflow:
  1. Annotate all objects on an image
  2. Hold Ctrl/Cmd and click each annotation to select them all
  3. Copy all annotations with Ctrl/Cmd + C
  4. Go to a similar image
  5. Paste all annotations at once
  6. Adjust individual positions as needed

​
Use Grid Guides for Alignment

Enable the grid guide (click the Guía button) to help align annotations consistently. This is especially useful for:
  • Architectural images with straight lines
  • Product photos requiring precise positioning
  • Creating annotations of consistent size
The grid is displayed at 32px intervals (configurable via GRID_SIZE constant) and helps maintain visual consistency across annotations.

​
Label Management

​
Bulk Label Creation

Use the bulk label textarea to create multiple labels at once. Enter labels separated by commas or line breaks:
car, truck, bus, motorcycle
person, bicycle
traffic light, stop sign
Click Agregar lote to create all labels in one operation.

​
Label Naming Conventions

Best practices for label names:
  • Use lowercase (labelWise automatically converts to lowercase)
  • Keep names short but descriptive (car not automobile_vehicle)
  • Use underscores for multi-word labels (traffic_light not traffic light)
  • Be consistent across your dataset
  • Avoid special characters and emojis

​
Pre-Define Your Label Set

Create all your labels before you start annotating. This prevents:
  • Typos and inconsistent naming
  • Having to switch between drawing and label creation
  • Accidentally creating duplicate labels with different names

​
Canvas Navigation

​
Zoom and Pan Effectively

For detailed work:
  1. Switch to Pan mode (click Mover button)
  2. Use the + button to zoom in up to 300%
  3. Drag the canvas to navigate
  4. Switch back to Draw mode to create annotations
  5. Use Reset to return to 100% zoom
Pro tip: Zoom in when working on small objects or fine details. The minimum box size is 10×10 pixels, so zooming helps with precision.

​
Two-View Workflow

Toggle between Canvas and CSV views to leverage different perspectives:
  • Canvas view: Visual annotation, drawing, positioning
  • CSV view: Precise numeric adjustments, bulk review, label changes
Use CSV view to:
  • Quickly change labels without clicking through the canvas
  • Enter exact pixel coordinates for perfect alignment
  • Review all annotations across all images in one table
  • Multi-select rows with Ctrl/Cmd + Click

​
Export Habits

​
Save Early, Save Often

Critical: Export your CSV frequently. labelWise runs entirely in your browser with no server-side storage. If you close the tab or refresh the page, all annotations are lost.
Recommended export schedule:
  • After completing each batch of 10-20 images
  • Before taking a break
  • Before closing the browser
  • After any significant annotation session

​
Incremental Exports

Exported CSV files are named annotations-YYYY-MM-DD.csv. To avoid overwriting:
  1. Rename downloaded files immediately (add batch number, timestamp, etc.)
  2. Keep a backup folder with dated exports
  3. Merge CSVs in Excel or Python if working in multiple sessions

​
Annotation Quality Tips

​
Consistent Bounding Box Strategy

Decide on a consistent strategy for all annotations: Tight boxes: Touch the object edges with minimal background
  • Pros: More precise, better for segmentation models
  • Cons: Time-consuming, harder on complex shapes
Loose boxes: Include some background margin
  • Pros: Faster annotation, more forgiving
  • Cons: Less precise, may include irrelevant context
Choose one strategy and stick to it across your entire dataset. Inconsistent bounding box styles can confuse ML models during training.

​
Handle Occlusions

For partially visible objects:
  • Annotate the visible portion only (don’t guess the hidden part)
  • Use consistent label names for partial views if relevant to your model
  • Consider creating separate labels like person_partial if occlusion is important

​
Edge Cases

Objects cut off by image boundary:
  • Annotate the visible portion
  • Extend box to image edge (coordinates are clamped automatically)
Overlapping objects:
  • Annotate all objects, even if boxes overlap
  • Z-order doesn’t matter in CSV export
  • Use multi-select to manage overlapping annotations

​
Performance Tips

​
Image Loading

For large datasets:
  • Load images in batches of 50-100 at a time
  • Remove completed images before loading the next batch
  • Use consistent image sizes to maintain predictable performance

​
Browser Considerations

  • Use Chrome or Edge for best performance
  • Close unused tabs to free memory
  • Disable unnecessary browser extensions
  • For very large images (>4K resolution), consider resizing before annotation

​
Collaborative Workflows

​
Splitting Work Across Team Members

  1. By image set: Divide images into folders, each person annotates their folder
  2. By label type: Person A annotates vehicles, Person B annotates pedestrians
  3. By review phase: Person A annotates, Person B reviews and corrects

​
Merging Annotations

Since labelWise uses CSV format, you can merge work: 
import pandas as pd

# Load multiple CSV exports
df1 = pd.read_csv('annotations-person-a.csv')
df2 = pd.read_csv('annotations-person-b.csv')

# Merge and deduplicate
merged = pd.concat([df1, df2]).drop_duplicates()
merged.to_csv('annotations-merged.csv', index=False)

​
Keyboard-First Workflow

For maximum speed:
  1. Pre-create all labels
  2. Select first label
  3. Load images
  4. Draw annotations on first image
  5. Ctrl/Cmd + C to copy all annotations
  6. Arrow key or click to next image
  7. Ctrl/Cmd + V to paste
  8. Adjust positions
  9. Repeat from step 6
  10. Export CSV every 20 images
This workflow can achieve 10-15 images per minute for repetitive annotation tasks with consistent object placement.

Build docs developers (and LLMs) love

Get started for freeTalk to us