Computer suggestion being ''too precise''

[jasonhernandez74]

@rupertclayton

Let’s imagine that CV is trained on a database of typical smartphone photos of spiders that have already been accurately IDed via microscopy (a big task). CV might then discover aspects of those photos that reliably predict particular genera and species, even though the photos don’t contain the microscopic detail the experts would need to key out those taxa. That’s because not every subtle character is contained in a species description or a key. Species differ in ways we haven’t yet recognized.

Even if that happened, there would still be the problem of getting those experts on board with it when they are used to thinking that the species can only be distinguished with microscopy.

[rupertclayton]

This line of computer vision research would be interesting in relation to cryptic species, which currently may only be distinguishable via DNA sequencing. I can imagine taking a series of good-quality photos of two such species from a range of angles and then setting machine learning on the task of finding ways to select the right species with better-than-chance accuracy.

If it’s successful, then the project would show that there is some visual difference, even if we haven’t yet detected what. If computer vision generally cannot get better than a chance result, then we truly do have distinct species that appear identical to observers.

But that’s all independent from the question of how precise an ID iNat’s CV algorithm should reasonably offer for a particular taxon group…

[dianastuder]

@wynand_uys has deliberately set out to ID our spiders alive. To find the living field marks.