Dataset Versions
v3
prod new
Generated on May 8, 2023
831 Total Images
View All Images
Dataset Split
Train Set 96%
801Images
Valid Set 3%
21Images
Test Set 1%
9Images
Preprocessing
Resize: Fit within 380x380
Auto-Adjust Contrast: Using Histogram Equalization
Filter Null: Require at least 80% of images to contain annotations.
Augmentations
Outputs per training example: 3
Rotation: Between -22° and +22°
Grayscale: Apply to 40% of images
Exposure: Between -25% and +25%
Blur: Up to 3.5px
Noise: Up to 10% of pixels