Dataset Versions

v1

2024-03-10 9:50pm

Generated on Mar 10, 2024

Dataset Split

Train Set 88%
1053Images
Valid Set 8%
100Images
Test Set 4%
49Images

Preprocessing

Resize: Stretch to 512x512

Augmentations

Outputs per training example: 3
Flip: Horizontal, Vertical
90° Rotate: Clockwise, Counter-Clockwise, Upside Down
Crop: 0% Minimum Zoom, 44% Maximum Zoom
Rotation: Between -45° and +45°
Shear: ±26° Horizontal, ±11° Vertical
Exposure: Between -29% and +29%
Blur: Up to 4.1px
Noise: Up to 4.81% of pixels
Cutout: 9 boxes with 21% size each