Flowers
560 photographs of flowers from seven species, projected by their image content.
Classes
- Bluebell 80
- Lilly Valley 80
- Snowdrop 80
- Sunflower 80
- Tigerlilly 80
- Tulip 80
- Windflower 80
The mount around each thumbnail is its class. Nothing else is encoded — the picture is the data, and the colour is only there so you can see whether the projection grouped the classes or not.
Nothing is encoded
Section titled “Nothing is encoded”Every other showcase in this section argues about a mapping: which measurement drives which part of the glyph, and whether the reader can decode it. Here there is no mapping to argue about. The item is a picture, and the only question left is the one this library exists for — is there room to draw it.
That makes the image case the cleanest statement of the problem. A thumbnail small enough not to overlap is a thumbnail too small to recognise, and shrinking until the overlaps go away destroys exactly the thing you came to look at. Switch Layout and the same 560 photographs become individually visible, at a size set by the grid rather than by the closest pair.
The mount around each thumbnail is the species. It is there to answer a question the pictures cannot: whether the projection grouped what a botanist would group.
Where the projection is honest, and where it is not
Section titled “Where the projection is honest, and where it is not”Sunflowers and bluebells separate cleanly — strong, unmistakable colour. The white flowers do not: snowdrops and lily of the valley sit in the same region, because a projection computed from image content is largely a projection of colour and shape, and both are small white flowers on green.
That is worth seeing rather than being told, and you can only see it once the images are big enough to compare side by side. On the raw projection that neighbourhood is a smear.
References
Section titled “References”Automated Flower Classification over a Large Number of Classes Conference on Computer Vision, Graphics & Image Processing 722–729, 2008. doi:10.1109/ICVGIP.2008.47
The flower categories these photographs come from.
Hagrid: using Hilbert and Gosper curves to gridify scatterplots Journal of Visualization 25(6), 1291–1307, 2022. doi:10.1007/s12650-022-00854-7
Where this example first appeared.