Dataset Versions

v4

Training with random weights- Resize-416x416-

Generated on Aug 13, 2024

Dataset Split

Train Set 88%
1659Images
Valid Set 8%
147Images
Test Set 4%
74Images

Preprocessing

Auto-Orient: Applied
Static Crop: 25-75% Horizontal Region, 25-75% Vertical Region
Resize: Stretch to 416x416
Auto-Adjust Contrast: Using Contrast Stretching
Filter Null: Require all images to contain annotations.

Augmentations

Outputs per training example: 3
Flip: Horizontal, Vertical
Crop: 0% Minimum Zoom, 20% Maximum Zoom
Rotation: Between -13° and +13°
Brightness: Between -15% and +15%
Exposure: Between -10% and +10%
Blur: Up to 4px
Noise: Up to 1.25% of pixels