High school photography teacher here, testing out iNat before trying it out with a few advanced student volunteers. The trouble is, I’m a photography and technology teacher, not a science teacher, and I’m no naturalist, alas, so we’ll be concentrating on taking quality observations, but then relying heavily on the AI and the community for the ID part.
My question: I experimented with taking a series of photos of a small evergreen tree outside our school - an overview, a closeup of the needles, a closeup of a cone, etc. The overview was the first shot, but it was a little busy, so the AI came up with “unknown”. (I understand that the AI only looks at the first photo?) Then I figured out how to reorder the photos, so that the cone shot was the first photo, instead.
…But how do I reactivate the AI to look at the (new) first photo to give it a second crack at IDing the series?
Once you re-order the photos, you can just click the same button you clicked to get the first computer vision generated ID. If that doesn’t help, perhaps you can share some screenshots so we can understand exactly what is happening.
Hey, welcome to iNat! It sounds like a great and fun project! To answer your question: It depends. Which platform are you using?
Do you upload the photos via the app (if so, Android, or iOS?) or the website?
One thing regarding the AI: Please don’t blindly trust the AI (or any single IDer for that matter) and try to verify any ID independently. Initially, broad manual IDs (like “Insect” or “Bird”) are usually better than a false overconfident ID and help the ID get to Research-Grade faster. :)
I have no experience with the app, but if you’re using the website, after uploading the various photos, you can use computer vision on each image before merging them into one observation. If the CV comes up with the same suggestion for each image, that’s a good indication that I’m moving in the right direction. But remember it will always be just a suggestion that will need verifying.
I’ll add that you don’t have to use the AI – if you have a general idea what it is (e.g., an evergreen) you can type in the name and the database will look for a match (in this case, order Pinales, conifers/evergreens) and you can select this as your ID.
I’m so glad you are trying it out! You’re going to get such better results for having worked out the bugs.
The suggestions and clarifying questions received so far are all spot on for my recommendations. I don’t have the app, don’t want the app (for a few reasons) and I upload all my photos using the website once I get home. I’ve heard that many functions of the iNaturalist are easier to find on the website.
I have two approaches for checking multiple photos.
DURING UPLOAD: I will upload all photos of one organism separately - so if I have 4 images of an insect, it looks like I’m making 4 observations (obs). I then look at the CV (Computer Vision - what the AI is called) suggestions. Do they agree over all four? Or do they disagree? If they disagree and I feel it’s due to the quality of the photo (too far away? too dark? occluded by a twig?), then I know I don’t want that one to be the first. If they disagree and I feel it’s because it’s really unsure, I might go with a higher suggestion (fly instead of blow fly or insect instead of blow fly) till I can check later.
If I’ve got the time, I might take this initial moment to so some research. What is the occurrence of the suggested ID in my area? Can reference materials help me decide once I’ve honed in on a genus or family level suggestion? But I often don’t feel like I’ve got loads of time at this point because there are hundreds of other photos waiting their turn!
AFTER UPLOAD: when I want to come back later to give a better try at a finer ID, I will use the edit function to reorder the photos. This might be available on the app. On the website, once you hit ‘Edit’, you’ll find the option to reorder under the thumbnails of the photos. It’s small text and easy to overlook. I’ll actually go in and do this multiple times seeing what results I get from each photo. Sometimes, I’ll download one of the photos, crop it tighter, and upload the cropped photo to let the CV have a stab at that.
I often tell people I’m trying to recruit to iNat to not worry about making mistakes. All mistakes can be corrected. I think mistakes will be made while learning how to use the platform. And I think people feel embarrassed that they didn’t get it right the first time and just say ‘forget it’. I don’t want that to happen. But, it’s important to know that ‘All mistakes can be corrected’ if one stays active on the site and checks their notifications and responds to comments and suggestions. If they upload, make a mistake, and then never come back - many of those mistakes can’t be corrected by other people.
I would love it if you could report back how well your project went and how much the students liked or didn’t like interacting with nature and using iNat as a support for interacting with nature.
Bear in mind that you don’t need to ID the observations at all. I mean, something basic like plant or bird is helpful, with conifer or duck being even better. Basically, though, if your students are posting good, identifiable photos with accurate dates and places, they’re making a useful contribution. We identifiers can take it from there.
If you’re relying heavily on the AI, just note to yourself and your students not to blindly echo the first person who contributes, either. If you don’t know what it is, agreeing with a random person and making it instantly research grade is not in anyone’s best interest. Once RG, it takes it out of the review pool.
I find the AI works best before the photo is uploaded. Once it’s public, it’s going to judge by the first photo. Before it’s public you can have it individually scan photos.
Your approach for the website works, but I personally find it pretty time-consuming and clunky.
An alternative approach is to use the mobile app on a smartphone (or a tablet probably as well) and then swipe through the photos for suggested IDs for each photo.
I wish the same could be done on the web application, but I’m guessing they’re either trying to reduce the computing loads, or haven’t wanted to get into coding that yet on the back end.