Dataset Versions

v9

2023-07-08 3:34pm

Generated on Jul 8, 2023

Dataset Split

Train Set 93%
4746Images
Valid Set 5%
260Images
Test Set 2%
99Images

Preprocessing

Resize: Fit within 640x640
Modify Classes: 2 remapped, 1 dropped
Filter Null: Require all images to contain annotations.

Augmentations

Outputs per training example: 3
Flip: Vertical
90° Rotate: Clockwise, Counter-Clockwise, Upside Down
Exposure: Between -10% and +10%
Blur: Up to 2.5px
Noise: Up to 3% of pixels