Dataset Versions

v72

2025-01-06 6:08pm

Generated on Jan 6, 2025

Dataset Split

Train Set 92%
657Images
Valid Set 4%
28Images
Test Set 4%
26Images

Preprocessing

Auto-Orient: Applied
Resize: Fit (reflect edges) in 960x960
Auto-Adjust Contrast: Using Histogram Equalization
Filter Null: Require all images to contain annotations.

Augmentations

Outputs per training example: 3
Flip: Horizontal
Crop: 0% Minimum Zoom, 10% Maximum Zoom
Rotation: Between -10° and +10°
Saturation: Between -15% and +15%
Brightness: Between -15% and +15%
Exposure: Between -10% and +10%
Noise: Up to 1% of pixels