Dataset Versions

v25

640-v2

Generated on Mar 23, 2024

Dataset Split

Train Set 94%
3304Images
Valid Set 4%
130Images
Test Set 2%
65Images

Preprocessing

Auto-Orient: Applied
Resize: Stretch to 640x640

Augmentations

Outputs per training example: 7
Crop: 0% Minimum Zoom, 30% Maximum Zoom
Rotation: Between -15° and +15°
Shear: ±15° Horizontal, ±15° Vertical
Hue: Between -25° and +25°
Saturation: Between -30% and +30%
Brightness: Between -20% and +20%
Exposure: Between -15% and +15%
Blur: Up to 1.5px
Noise: Up to 1% of pixels