Dataset Versions

v2

2024-10-07 4:08am

Generated on Oct 7, 2024

Dataset Split

Train Set 87%
1104Images
Valid Set 9%
112Images
Test Set 4%
48Images

Preprocessing

Auto-Orient: Applied
Static Crop: 25-75% Horizontal Region, 25-75% Vertical Region
Resize: Fit within 640x640
Tile: 4 rows x 4 columns

Augmentations

Outputs per training example: 3
Flip: Horizontal, Vertical
90° Rotate: Clockwise, Counter-Clockwise, Upside Down
Brightness: Between -15% and +15%
Exposure: Between -10% and +10%