In the web app to confirm/improve the IDs of others I have tried using the “Suggestions” tab but it rarely does well in a case where the picture is ambigous or maybe contains a couple of species. What does work really well is to use a print screen/snipper app to zero in on the species believed to be the one the observer wanted and then upload the screen print as a separate test observation. Suggestions only seems to work well when the answer is obvious. This is particularly strange when trying to identify unknowns following the City Nature Challenge. When something was obviously a plant to me, the suggestion always came up as a frog/toad whereas the print screen might get me in the ball park.
if anyone were to be so inclined, i think it would be technically possible to write a browser extension that would allow the user to select a square crop from an image in iNaturalist, send a standard-sized version of the crop to the computer vision API, and report the suggestions to the user. as long as the requests were done from an iNaturalist page, i think it would work.
someone could also make a feature request to add similar functionality to the system.
5 Likes
I use Google Lens for the iffy obs - then I can crop down to This - is - A Spider!
1 Like
I haven’t used Google Lens. Does it look at the iNaturalist page or do you need to cut and paste into a separate browser window? Or are you pointing your phone at the image in the web browser?
On the website. Presuming you are using Google, rightclick on the photo, then Google Lens. No cut, no paste. The option to crop is there. It searches for info across the internet - which often includes iNat. I ID Unknowns so I sometimes need help working out what even IS that?
If I can start with a broad ID on iNat - that often helps to trigger CV to offer more useful options. And I can use Checklist or Visually Similar options.
different browsers will implement their own visual search thing for images in webpages. in Chrome, you can right-click on an image and select Google Lens, as dianastuder noted, but other browsers might implement it a little differently.
1 Like
What I do (on my computer) is download the photo, crop the image, then submit the cropped photo to the Computer Vision Demo.
2 Likes