ISBN Visualization

This picture is 65,536×32,768 pixels. Each pixel represents a single ISBN. If the ISBN is contained in Anna's Archive (AA), the pixel is blue. If we know that an ISBN has been issued, but we don’t have a matching file, we make it red. A white pixel means that the ISBN is unknown by AA or does not exist.

Features

• The map visualizes all ISBN-13 identifiers (978 and 979 prefixes) using a Hilbert curve, which is a space-filling fractal that maps a 1D sequence (in this case, all ISBNs in increasing order) onto a 2D space while preserving locality. This means that ISBNs which are numerically close together will also be clustered together in this map. As ISBNs are assigned as continuous ranges to countries and publishers, interesting patterns and structures emerge.
• For the combined dataset, the darker the red, the more datasets contain the ISBN (but AA doesn't). For the combined_inv dataset, the opposite is true, the darker the red, the less datasets contain the ISBN (again AA doesn't). In both cases, blue means that AA has the ISBN linked to an archived book.
• In addition to the base dataset, an overlay dataset can be toggled which allows to compare datasets by changing the opacity of the overlay
• The country/language group to which a given ISBN belongs can be shown as a polygon vector overlay.
• When clicking on a pixel, the ISBN, country/lanuage, and a link to Anna's Archive are presented.
• On the frontend, the visualization of each dataset is done via image tiling using Openlayers and a statically hosted zoomify folder. This ensures that the original 65,536×32,768 (138.9 MB) does not need to be loaded all at once but is split up into 172 4096x4096 tiles (total 178,5 MB, biggest single tile 10 MB).

Credits

• Anna's Archive for the datasets, hosting the contest and overall support.
• OpenLayers for a really nice mapping library with lots of great examples.
• PNGJ for a high performance and large-size PNG library. The original combined_inv.png contains 2,147,483,648 pixels which is just one too much for Java Image I/O (which was also very slow).
• libvips/pyvips for high performance image processing, allowing me to transform a PNG into zoomify tiles.
• Lots of other awesome open source software that I will continue to link here once I parse through the code I wrote.
• Cloudflare Pages for its very generous free tier.
• ChatGPT, Claude.ai, Deepseek and Cursor

Source

github.com/charelf/isbnviz developed during January 2025

Contest update

On 2025-02-24 AA announced the winners of the contest and I am very happy to announce that I won a shared 3rd place 🎉! I had a lot of fun with the computational challenges and then the visualisation of the information as it combined several areas of interest for me. Thank you Anna's Archive for organising this contest!