README.md

# OCR demo files: Food for Space Flight

Files for trying OCR on NASA's *Food for Space Flight* booklet.
The [matching dataset](https://huggingface.co/datasets/uv-scripts/ocr-demo)
contains seven page images with source-page metadata.

| Path | Contents |
|---|---|
| `original/food-for-space-flight.pdf` | The unmodified nine-page NASA PDF. |
| `demo/food-for-space-flight.pdf` | PDF pages 3-9, matching the seven dataset rows. |
| `pages/` | The same seven page images and metadata. |

Use one prefix as your input: `hf://buckets/uv-scripts/ocr-demo/demo` for the
seven-page PDF, `/original` for the complete document, or `/pages` for images.
Pointing a recursive OCR recipe at the Bucket root would process the same pages
in multiple formats.

## Source and licence

Source: NASA, *NASA Facts: Food for Space Flight*, NTRS document 19830005543.

- [NASA source record](https://ntrs.nasa.gov/citations/19830005543)
- [Original PDF](https://ntrs.nasa.gov/api/citations/19830005543/downloads/19830005543.pdf)
- [Document metadata](https://ntrs.nasa.gov/api/citations/19830005543)
- [NASA usage guidance](https://www.nasa.gov/nasa-brand-center/images-and-media/)

NASA's document record states **Work of the US Gov. Public Use Permitted**.
The source pages retain that status; the repository's code licence does not
relicense the document. Attribute the source pages to NASA. NASA does not endorse
this demo or verify any generated OCR output.

Pages 3-9 (printed pages 2-8) were rendered at 300 DPI and saved as JPEG at quality
85. `manifest.json` records the source URLs, file sizes and SHA-256 checksums.
The archival scan has damaged lettering and an imperfect existing OCR layer;
this is demo input, with no verified ground-truth transcription.