Dataset Versions

v3

2023-05-23 3:35pm

Generated on May 23, 2023

Dataset Split

Train Set 85%
2391Images
Valid Set 8%
218Images
Test Set 7%
210Images

Preprocessing

Auto-Orient: Applied
Resize: Stretch to 640x640
Modify Classes: 3 remapped, 0 dropped

Augmentations

Outputs per training example: 3
Flip: Horizontal, Vertical
90° Rotate: Clockwise, Counter-Clockwise, Upside Down
Rotation: Between -27° and +27°
Shear: ±36° Horizontal, ±2° Vertical