Dataset Versions

v2

2023-09-02 12:20am

Generated on Sep 1, 2023

Dataset Split

Train Set 92%
19806Images
Valid Set 7%
1403Images
Test Set 1%
255Images

Preprocessing

Auto-Orient: Applied
Resize: Stretch to 640x640
Modify Classes: 26 remapped, 0 dropped

Augmentations

Outputs per training example: 3
Flip: Horizontal
Crop: 0% Minimum Zoom, 20% Maximum Zoom
Shear: ±15° Horizontal, ±15° Vertical
Saturation: Between -25% and +25%
Blur: Up to 2.5px
Cutout: 3 boxes with 10% size each

Similar Projects

See More