Augmentation Controls
Build, test, and randomise a pipeline.
Advanced effects are simulated in-browser and may alter labels.
Interactive Canvas
Drag, scale, rotate, and inspect the active image.
760×560
Canvas size
0
Transformations
0%
Diversity score
100%
Quality score
100%
Label preservation
0
Moves used
Original Image
Transformation History
No transformations yet.
Transformation Pipeline
Enable, reorder, duplicate, or remove steps.
Generated Batch
Create multiple variations from the active pipeline.
Challenge Mode
Complete practical augmentation missions.
Balanced Starter
Use at least three label-safe transformations and keep quality above 70%.
Time: 02:00 · Budget: 6 moves
Challenge score0
Dataset Expansion
5,000
Expanded training images
Performance Dashboard
Achievements
First Transform
Pipeline Builder
Diversity 70
Label Guardian
Batch Creator
Learning Guide
Label-safe: preserves the training label for most tasks.
Risky: may change object meaning, orientation, or visibility.
Generalisation: useful variation helps models handle unseen images.
Data leakage: never augment validation or test images using information learned from the complete dataset.