Dataset Versions

v4

2022-10-29 5:40pm

Generated on Oct 29, 2022

Popular Download Formats

Pascal VOC XML
Common XML annotation format for local data munging (pioneered by ImageNet).
PaliGemma
PaliGemma JSONL format used for fine-tuning PaliGemma, Google's open multimodal vision model.
CreateML JSON
CreateML JSON format is used with Apple's CreateML and Turi Create tools.
Other Formats
Choose another format.

Dataset Split

Train Set 87%
3675Images
Valid Set 8%
352Images
Test Set 4%
189Images

Preprocessing

Auto-Orient: Applied
Isolate Objects: Applied
Resize: Stretch to 640x640
Modify Classes: 0 remapped, 2 dropped
Filter Null: Require all images to contain annotations.

Augmentations

Outputs per training example: 3
Flip: Horizontal
90° Rotate: Clockwise, Counter-Clockwise
Crop: 0% Minimum Zoom, 70% Maximum Zoom
Rotation: Between -20° and +20°
Shear: ±15° Horizontal, ±15° Vertical
Hue: Between -45° and +45°
Saturation: Between -50% and +50%
Brightness: Between -50% and +0%
Exposure: Between -25% and +25%
Blur: Up to 1px
Noise: Up to 5% of pixels
Cutout: 10 boxes with 3% size each
Bounding Box: Flip: Horizontal, Vertical
Bounding Box: 90° Rotate: Clockwise, Counter-Clockwise, Upside Down
Bounding Box: Crop: 0% Minimum Zoom, 10% Maximum Zoom
Bounding Box: Rotation: Between -15° and +15°
Bounding Box: Shear: ±10° Horizontal, ±10° Vertical
Bounding Box: Brightness: Between -30% and +30%
Bounding Box: Exposure: Between -27% and +27%
Bounding Box: Blur: Up to 1px
Bounding Box: Noise: Up to 5% of pixels