Dataset Versions

v4

2024-08-28 9:33pm

Generated on Aug 28, 2024

Dataset Split

Train Set 94%
834Images
Valid Set 4%
38Images
Test Set 1%
13Images

Preprocessing

Auto-Orient: Applied
Resize: Stretch to 640x640
Auto-Adjust Contrast: Using Histogram Equalization

Augmentations

Outputs per training example: 3
Flip: Horizontal, Vertical
90° Rotate: Clockwise, Counter-Clockwise, Upside Down
Rotation: Between -25° and +25°
Shear: ±32° Horizontal, ±14° Vertical
Blur: Up to 1.3px
Cutout: 4 boxes with 14% size each