data/ufo_mnist_v1/dataset_card.md

# UFO-MNIST Dataset Card

UFO-MNIST is a deterministic MNIST-style dataset for classifying low-resolution
UFO-like shapes, sighting patterns, and common lookalikes.

## Intended Use

This dataset is for demos, tutorials, baseline model experiments, and visual
classification exercises built around common UFO/UAP spotting categories.

## Composition

- Image size: 28x28
- Color: grayscale
- Pixel dtype: uint8, range 0-255
- Classes: 10
- Images per class: 1000
- Train per class: 800
- Test per class: 200
- Total images: 10000

Public UFO/UAP sighting references, official release material, and permissive
public-domain records are used as visual and provenance references. The final
28 by 28 arrays are balanced with generated and augmented examples so every
class has the same number of train and test samples. Legally ambiguous public
galleries are excluded by default.

## Labels

- 0: `disk`
- 1: `orb`
- 2: `triangle`
- 3: `cigar_rod`
- 4: `light_formation`
- 5: `irregular_glow`
- 6: `aircraft`
- 7: `balloon`
- 8: `bird`
- 9: `celestial_or_artifact`

## Limitations

The labels are approximate shape categories. The generator deliberately adds
noise, blur, compression-like artifacts, and background variation to match the
low-resolution spotting style while keeping the dataset compact and reproducible.