Dataset Versions

v2

2024-02-25 1:32pm

Generated on Feb 25, 2024

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 98%
18009Images
Valid Set 1%
203Images
Test Set 0%
76Images

Preprocessing

Auto-Orient: Applied
Isolate Objects: Applied
Static Crop: 10-90% Horizontal Region, 10-90% Vertical Region
Resize: Stretch to 200x200
Auto-Adjust Contrast: Using Contrast Stretching
Grayscale: Applied
Tile: 2 rows x 2 columns
Filter Null: Require at least 50% of images to contain annotations.

Augmentations

Outputs per training example: 32
Flip: Horizontal
90° Rotate: Clockwise
Crop: 12% Minimum Zoom, 71% Maximum Zoom
Rotation: Between -45° and +45°
Shear: ±45° Horizontal, ±45° Vertical
Grayscale: Apply to 50% of images
Hue: Between -180° and +180°
Saturation: Between -50% and +50%
Brightness: Between 0% and +91%
Exposure: Between -30% and +30%
Blur: Up to 1.5px
Noise: Up to 50% of pixels
Cutout: 24 boxes with 71% size each
Mosaic: Applied
Bounding Box: Flip: Horizontal
Bounding Box: 90° Rotate: Clockwise
Bounding Box: Crop: 12% Minimum Zoom, 71% Maximum Zoom
Bounding Box: Rotation: Between -45° and +45°
Bounding Box: Shear: ±45° Horizontal, ±45° Vertical
Bounding Box: Brightness: Between 0% and +91%
Bounding Box: Exposure: Between -30% and +30%
Bounding Box: Blur: Up to 1.5px
Bounding Box: Noise: Up to 50% of pixels