Dataset Versions

v1

2024-03-07 1:23pm

Generated on Mar 7, 2024

Dataset Split

Train Set 87%
5268Images
Valid Set 8%
502Images
Test Set 4%
251Images

Preprocessing

Auto-Orient: Applied
Resize: Stretch to 640x640

Augmentations

Outputs per training example: 3
Rotation: Between -10° and +10°
Hue: Between -17° and +17°
Saturation: Between -40% and +40%
Brightness: Between -35% and +35%
Exposure: Between -18% and +18%
Blur: Up to 2.5px
Noise: Up to 0.1% of pixels
Cutout: 18 boxes with 2% size each