Dataset Versions

v2

2023-06-30 9:07am

Generated on Jun 30, 2023

Dataset Split

Train Set 72%
833Images
Valid Set 22%
257Images
Test Set 6%
73Images

Preprocessing

Auto-Orient: Applied
Resize: Fit (reflect edges) in 2048x1365
Modify Classes: 66 remapped, 0 dropped

Augmentations

Outputs per training example: 3
Flip: Horizontal, Vertical
90° Rotate: Clockwise, Counter-Clockwise, Upside Down
Exposure: Between -12% and +12%