Matching Plant Images Internally

[Jorge_Stolfi]

What do you have against f5v?

The "tennis racket" stem.  Ditto for f40r.  For f11r, the problem is the leaves with two stalks going to distinct branches.

One might think that the illustrators were sitting next to each other.

Hard to believe since each herbal page must have taken as long to draw as a whole Pharma page with a dozen plants.  

Besides it would not explain why You are not allowed to view links. Register or Login to view. and f102r1-SE-corner have precisely the same perspective view of the root.  And yet the 3d shapes of the root are quite different.

Basically, 80% of the VM plants are authentic and known as medicinal plants.

  

Yes, some of the plants are partly copied from nature or from more or less "honest" herbals. I myself claim the glory of having identified conclusively You are not allowed to view links. Register or Login to view. as the pansy or forget-me-not (Viola tricolor), a common garden plant.  Not just because of the flower shapes (which are characteristic but upside-down), but because of the two leaf shapes: simple lances near the base, but changing to multiple fingers higher up.

But that fact does not dispel the thesis that the Herbal plants are based on Pharma-like sketches of some plant parts, like those of Pharma, with the rest being made up by the Scribe with the help of other herbals.  

And it does not imply that a plant whose drawing (whole or part) can be identified is the topic of the respective page.   Is the text on You are not allowed to view links. Register or Login to view. about Viola tricolor, or about some other plant that has a vaguely similar root?  I would not bet on the former...

And finally beware that there are millions of plant species, with more different leaf and flower shapes than one can imagine.  Thus any drawing of a plant, even one drawn by a five-year-old, can be matched to some species -- within the tolerance that people have been using in this task.  

It is like trying to read the text on page f116v: if you broaden the range of acceptable discrepancies until you get some match, you also get dozens of other possible matches.  And probably all wrong.

Take for example the plant of You are not allowed to view links. Register or Login to view. and f102r1-SE-corner.  The shape of the leaves is a key criterion for species identification, but it is significantly different in the two drawings. That shows that the plant drawings have errors of at least that magnitude.  So, which of the two shapes should be look for?  Or should we accept any other shape that differs from both as much as they differ from each other?

Or take the plant on f2r.  Everybody agrees that it is some species of Centaurea.  And indeed the flowers and calyxes are a rather good match -- especially since there are species with all possible petal shapes.   But what about the leaves?  The drawing has large somewhat broad leaves in compact bundles of five overlapping leaves.  I could not find an image of actual Centaurea whose leaves even remotely resemble that.  Can you?

And then there is the "gnome with claws" root.  I suppose that, with enough imagination, one can consider it You are not allowed to view links. Register or Login to view..  But the error tolerance would have to be set so high than an elephant would match a cow...

All the best, --stolfi