Dataset Versions

v1

2024-01-03 2:07pm

Generated on Jan 3, 2024

Dataset Split

Train Set 87%
180Images
Valid Set 10%
20Images
Test Set 4%
8Images

Preprocessing

Auto-Orient: Applied
Static Crop: 25-75% Horizontal Region, 25-75% Vertical Region
Dynamic Crop: Class: surface_scratches
Resize: Stretch to 640x640
Tile: 2 rows x 2 columns

Augmentations

Outputs per training example: 3
Flip: Horizontal, Vertical
90° Rotate: Clockwise, Counter-Clockwise, Upside Down
Crop: 0% Minimum Zoom, 22% Maximum Zoom
Rotation: Between -14° and +14°
Hue: Between -12° and +12°
Brightness: Between -23% and +23%
Noise: Up to 0.97% of pixels
Cutout: 3 boxes with 5% size each
Bounding Box: Exposure: Between -2% and +2%

Similar Projects

See More