Does the CV work differently on the desktop than it does on the Android phone app? Today an observation showed that the observer noted as Family Trichiaceae. On my desktop, the system said “No Suggestions Available.” When I tapped “Suggest ID” for it on my phone, the system returned several, starting with Trichiaceae and including a number of suggestions that looked plausible (to me - no expert in slime molds). In fact, I don’t think I’ve ever seen the phone return no suggestions, while the desktop periodically returns no suggestions, or a random list (maybe the most frequently submitted taxa that day?).
Are there in fact 2 systems? Inquiring minds want to know ;-)
I’ve noticed some discrepancies as well and been trying to quantify them. My experience (caveat - I’ve only been using it a few months) has been the “Suggest an Identification” box in the observation seems to correlate closely with the suggestions on the phone (iPhone in my case). The suggestions that come up in the suggestions tab of the identify mode or that come up when you hit compare seem to be different - they seem more correlated with what’s found in the area.
The default “Compare” or “Suggestions” page does not use computer vision, but rather displays a list of other closely related species that have been found in the region of the observation. You have to manually change it to “Visually Similar” in the filters and clear the taxonomic filter in order to use computer vision in the same way from that tab:
As of right now we are using the same model everywhere. The model has a web-based version (used by the website and iNaturalist apps) and a mobile version (used by Seek). The two versions are from the same training run, but can produce slightly different results because of compression in the smaller [Seek-only] mobile version, and because the web-based version can weight results based on nearby species occurrences.
The default “Compare” or “Suggestions” page does not use computer vision, but rather displays a list of other closely related species that have been found in the region of the observation. You have to manually change it to “Visually Similar” in the filters and clear the taxonomic filter in order to use computer vision in the same way from that tab…
Thanks! That explains it. I didn’t realize I could toggle the Source - very useful!
All useful info - thanks. I may well be too impatient to wait for it to load. I also had forgotten about the Seen Nearby bit. I’ll do some more testing and see if I can confirm patterns (after waiting for a full load!).
