Dataset Versions

v2

2023-07-22 8:04pm

Generated on Jul 22, 2023

Dataset Split

Train Set 82%
19712Images
Valid Set 12%
2816Images
Test Set 6%
1408Images

Preprocessing

Auto-Orient: Applied
Static Crop: 9-91% Horizontal Region, 5-96% Vertical Region
Grayscale: Applied
Tile: 4 rows x 4 columns
Modify Classes: 1 remapped, 0 dropped

Augmentations

Outputs per training example: 2
Shear: ±15° Horizontal, ±15° Vertical
Brightness: Between -25% and +25%
Blur: Up to 2.5px
Noise: Up to 5% of pixels