When I’m uploading large batches of observations using the desktop site, I don’t often want to take the time to do ID for every picture. The problem is, this seems to be the only time I can make the AI give me ID suggestions for each picture–ie, before I group them into a single observation. I like to put several pictures for each observation, and since different angles give different suggestions, you can sometimes “triangulate” a much better guess than you get from one picture alone.
At the moment, the only way I know to do this is to literally open the Add Observations tab again and add the pictures again. And of course I can only do that for my own observations. Is there a way to do this that I’m not aware of? Or is it a feature that could be added?
I feel your pain since I run into this a lot with identifying other people’s observations. Sometimes, the second photo for instance really looks familiar and if I had a clue or two from the AI then I could nail it down, but all you get are suggestions from a less-notable first picture.
I upload from mobile using my phone’s camera feature and assemble each observation by selecting from the phone’s gallery. I have been consciously selecting what I think might be “the easiest AI picture” as the first one in my obs in case others will have the same issue.
You should be grouping photos of the same organism (or different views/angles) as a single observation. The AI will take into account all photos, as far as I am aware.
I’m not so sure of that, if it takes them all into account, they’re certainly weighted differently! I have changed the order of the photos on my own observations and the suggestions given by the computer vision change when I do so.
I’ve always wondered how the website selects which photo to display first. You would think they’d be in the order taken, or the order added to the uploading screen, but neither seems to be true. It’s very annoying to have to re-order the photos on so many of my observations, and the method of doing so seems clunky–it takes at least three clicks and manually typing in numbers.
Based on what lvdt is seeing, I suspect the computer vision is only matching against the first photo. I have to imagine that matching against more than the first photo would be a coding nightmare. When I can, I reorder in the app to put what I think might be the best photo first and, if I do not know the plant myself, use the computer vision after that resequencing. Even then the computer vision must be used with care and attention to possible range issues.
When you drag and drop photos to the uploader, the last one clicked will be the first one dropped. So if you have 5 photos in your harddrive folder, click the first, hold shift and click the last to bulk drag them, then the “drop order” will be 5 1 2 3 4. If you click #5 and then shift-click #1, the "drop order will be 1 2 3 4 5. If you make the selection in any order, and then click and drag say #4, then all 5 are dragged and the "drop order will be 4 1 2 3 5.
Once into the uploader, the same behaviour occurs, but what you drop is on top of something else, which becomes the first.
This is a standard windows behaviour, and occurs when dragging and dropping in other windows applications eg file manager.
The thing that makes this more than slightly inconvenient is that you can’t use the compare tool in the uploader. So in order to apply the full functionality of the CV to the second or third photo on an observation, I either have to reorder the pictures or actually submit that picture as a unique observation (and then delete it). Based on the responses I’m getting, it sounds like this is not something the system is already set up to do? Should we make it a feature request? How do I do that?
@megachile Is there a way to make the AI give suggestions for the second or third picture in an existing observation?
A few ways:
- in the web uploader by running computer vision before combining the photos into one observation,
- split up the photos if you already combined them in the web uploader — click and drag the photo off of the observation card to split it up
- if you already uploaded them, change the order of the photos via the web observation edit interface
- if you already added the photos via mobile, first select an alternative default photo
If it’s not your own observation, you have to download the 2nd/3rd/etc image and use the computer vision demo page or upload page to use computer vision on an image other than the first one.
It sounds like you might have a few different requests here. One might be adding the Compare tool to the uploader while another would be some sort of way to use computer vision on additional images rather than just the first one. You can start a new topic in the Feature Requests category. More info here: https://forum.inaturalist.org/t/about-the-feature-requests-category/69/3
This topic was automatically closed 60 days after the last reply. New replies are no longer allowed.