This question came up recently. Is the first photo the one that best helps determine the ID? If someone posts a cropped photo of a leafmine and then posts additional photos showing large portions of the leaf or plant, do subsequent photos help or do they also create a problem with identification.
I guess my basic question is should a cropped photo be the first photo if there are multiple photos for leaf mines.
Only the first photo is taken in account for CV. In terms of the rest of the photos, you can add whatever you think would be helpful for identifiers, cropped or uncropped
I’m not sure what you mean by “best helps determine the ID”.
For training the CV, any of the photos in an observation may be used, not just the first one.
For the suggestions provided on observations, it depends: In the web interface, the CV only looks at the first photo. In the apps you can swipe between images and get suggestions for each one.
I find it helpful as an IDer if the first image is cropped or labelled in such a way that it is clear what the object of interest is. The first photo primes us about what to expect, so if it appears to show one thing but the observer is interested in something else, there is a risk that IDers may misinterpret the image and overlook any notes included by the observer.
A workaround to get the same functionality from the web interface (assuming you are the observer): ask for suggestions before combining the photos.
I was under the impression that only the first five photos in any observation are eligible for inclusion for CV training. According to this help page, only five photos per observation are included for model training.
I’m pretty sure we’re not just choosing from the first five, they could be any photos from the observation.
A post was merged into an existing topic: Link observations