Extracting ID's from computer vision upload

[yashsondhi]

Hi,
Is there a way to get the Computer vision ID’s to work on a batch of photos and extract the ID’s without having to submit them as observations? I am working on a project where I want to count the number of different insects orders that appear on a moth screen and have got cropped images of each insect, if I manually upload them to the species upload page, only the first image gets an ID, I have to go to each other image and hover over the suggested ID to get it. Is there some way to get this information using the API or in a manner that allows me to process more images together?

[fluffyinca]

I don’t really understand what your problem is. Are you saying that using the website uploader takes too long because you have to ID them one at a time? I think that’s the only way to use the CV, although I don’t know anything about the app. Also, why don’t you want to submit them as observations? Does your project need to be finished quickly?

[pisum]

iNat staff say:

@tiwane

There is no public API for iNaturalist’s vision model.

and

@kueda

if you’re referring to our vision model, we don’t distribute that

…

it sounds like you may want to develop your own computer vision model, go with one that’s publicly available, or just go without computer vision altogether. iNaturalist’s computer vision is not always great for arthropods in general. getting things down to the order level should be relatively straightforward for humans relatively familiar with insects. so it might be faster to just have an insect person go through a set of photos than to do anything with computer vision, depending on how many photos you have.

[bobmcd]

I am not sure I totally understand. I if I use the uploader and load multiple images (last time I did about 40 at once), in this case label their location in bulk by selecting all, I can get a suggested ID from the computer vision for each one. I can then choose not to submit and cancel. I have done this to help identify organisms for others while not submitting them as my own. You would then have to figure out a way to label the original image - you could do this with a copy paste rename of the original file and then add a number as part of the file name.