Dataset Versions

v3

2024-02-06 9:22am

Generated on Feb 6, 2024

Dataset Split

Train Set 83%
96Images
Valid Set 12%
14Images
Test Set 5%
6Images

Preprocessing

Auto-Orient: Applied
Resize: Fit within 224x224

Augmentations

Outputs per training example: 3
Flip: Horizontal, Vertical
90° Rotate: Clockwise, Counter-Clockwise, Upside Down
Saturation: Between -30% and +30%
Exposure: Between -15% and +15%
Blur: Up to 1.8px
Noise: Up to 1.49% of pixels
Cutout: 8 boxes with 11% size each