Dataset Versions

v3

2023-09-22 2:16pm

Generated on Sep 22, 2023

Dataset Split

Train Set 95%
2889Images
Valid Set 4%
137Images
Test Set 1%
24Images

Preprocessing

Auto-Orient: Applied
Resize: Stretch to 1024x1024

Augmentations

Outputs per training example: 3
Flip: Horizontal, Vertical
90° Rotate: Clockwise, Counter-Clockwise, Upside Down
Rotation: Between -15° and +15°
Saturation: Between -20% and +20%
Brightness: Between -15% and +15%
Blur: Up to 2.5px
Noise: Up to 2% of pixels
Cutout: 2 boxes with 7% size each