Dataset Versions
v6
2024-01-19 9:48pm
Generated on Jan 19, 2024
8429 Total Images
View All ImagesDataset Split
Train Set 57%
4774Images
Valid Set 25%
2074Images
Test Set 19%
1581Images
Preprocessing
Auto-Orient: Applied
Resize: Stretch to 640x640
Modify Classes: 14 remapped, 34 dropped
Filter Null: Require all images to contain annotations.
Augmentations
Outputs per training example: 2
Grayscale: Apply to 25% of images
Hue: Between -6° and +6°
Brightness: Between -15% and +15%
Blur: Up to 0.7px