Apologies if this has been discussed previously but I couldn’t find it in the other CV threads. Sometimes, observers will upload a single photo multiple times to create observations for each species shown in the photo (e.g., a mixed flock of shorebirds).
I’m curious about how this affects the computer vision’s “learning,” because I find myself second-guessing whether it’s a good idea to confirm some of these observations to bring them up to Research Grade. If a photo shows Species A prominently in dead center, and Species B is tucked off in a corner and maybe a bit out of focus, is elevating the photo to RG for Species B going to “confuse” the CV system? Same thing for photos where there’s 20 individuals of Species A and one individual of Species B mixed in.
Worrying about computer vision training seems utterly foolish (unless you’re one of the developers). If it can’t handle multiple species in one photo it’s useless. In fact, those photos are probably more useful.
I post photos from scuba diving and there are almost always multiple species in a photo. I try to include a crop or two of just the species I want to ID with a group shot to show scale and habitat. Computer Vision will just have to learn.
On the other hand, I duplicate the observation so the same photograph is used for both observations, rather than uploading the same photo twice. I’m not it makes a difference to Computer Vision training, but I feel better about database space and I like being able to find all the identifications from the one photo.
If i recall correctly the algorithm actually ‘looks’ at different parts of the photo and recognizes there is more than one organism… so definitely not a problem.
As long as the species selected is present in the photo, there is no reason not to confirm it to research grade. Users are not required to provide perfect photos, or cropped photos or anything else. They are required to provide evidence. Whether that evidence is front and centre in the photo, or tucked away in the corner, or one bird within a flock of many birds etc is immaterial to its eligibility for research grade status.