Dataset Versions

v2

2023-10-16 4:12pm

Generated on Oct 16, 2023

Dataset Split

Train Set 87%
1698Images
Valid Set 8%
162Images
Test Set 4%
81Images

Preprocessing

Auto-Orient: Applied
Resize: Stretch to 640x480

Augmentations

Outputs per training example: 3
Flip: Horizontal, Vertical
90° Rotate: Clockwise, Counter-Clockwise, Upside Down
Rotation: Between -3° and +3°
Saturation: Between -20% and +20%
Brightness: Between -5% and +5%
Bounding Box: Noise: Up to 10% of pixels