Amphiprion ocellaris and Amphiprion percula (AKA Nemo) are essentially the same fish, found in neighboring parts of the West Pacific. The only species present in the Great Barrier Reef is A. percula… and, yet, the Computer Vision AI fails to suggest this as the favored ID (or even as an option for the ID) in this observation. Can someone involved with this system explain how an error like this is possible? These fishes are, again, visually indistinguishable (for the most part) and best identified by biogeography. Why am I seeing several options for other less-similar congeners, but not for the nearly identical one that is actually found in this region?!
I’ve noticed this trend regularly, where obvious ID choices are passed over for nonsensical ones. With Giant Clams, of which there are ~10 species, I’ll often see just a few given as choices, with nudibranchs and octopodes often filling out the remainder of the suggested ID list (like with this observation, though this is probably applicable to every Tridacna observation). This seems like an excellent way to give users the false impression of surety. If you were identifying that Amphiprion and only saw A. ocellaris in the ID list, it wouldn’t be obvious that there was a better choice.
Do better, Computer Vision.