Many of our users rely on the the default taxon photo when selecting an ID from the list of suggestions. To give them the best chance of choosing the right species, it’s important that those photos are the best representation of the species that we have available. The additional curated photos can help users decide between similar species.
I was surprised today to discover that Fremont cottonwood, a species with hundreds of observations, had single, poor quality default photo. I try to curate photos as I run across ones with few or no photos but there are many taxa that need attention. The website pages show up to 5 photos and the apps let users see up to 12 photos so we should be using them.
Here’s my process for selecting photos, I would love to hear feedback from others. (Below primary=the default photo, secondary=the other curated photos)
Make sure the photos are the right species (I see a lot of flags about default photos of the wrong species.) Just because an observation is Research Grade, doesn’t mean it’s identified correctly. If you’re not familiar with the species, take a minute to do a little research before selecting photos. When the best photo is a cultivated specimen, make sure it’s close to the wild type, not a cultivar or hybrid.
It’s best to use internal iNat photos when possible. External websites change or the photos may not display at the same resolution. Many Flickr photos are not identified correctly and it does not offer the same peer review process as iNat.
Choose a primary photo that is clear and shows the most identifiable parts of the plant. Show the flowers but avoid extreme closeups of a single flower for the default photo. Closeups are good for the secondary photos but the default should show what a person would see from roughly arm’s length for most plants.
Choose a primary photo that is the most typical form for a taxon. If it usually has purple flowers, don’t choose an unusual white morph, save that for the secondary photos. If the taxon is variable, the secondary photos should reflect that. Show the variety of flower colors, leaf colors, heights, growth forms, etc.
Secondary photos should show other parts that will help a person identify their observation. Leaves, sepals, side of the flowers, habit, and anything special about the species is helpful. Remember that the web pages only show the first 4 secondary photos so put the important ones near the beginning.
Avoid photos with lots of other plant species. People who are unfamiliar with the species may have trouble distinguishing between which parts belong to the taxon on the page and which belong to other species in the photo.
Avoid photos with weird lighting, such as heavy contrast, backlighting, over or under exposed, or skewed white balance. You may realize that the color is off because of a deep shadow but someone glancing through thumbnails won’t.
Avoid using photos with long, narrow cropping for the default. Make sure it looks good at the thumbnail size and square crop.
Pollinators and other plant visitors should be reserved for the secondary photos.
Sometimes a person’s hand is useful for scale or holding a plant part to get a better angle but I try to avoid hands and feet in the primary photo.
If a taxon has a lot of observations, I usually scroll through the list and grab any photos that look promising, then weed out the lower quality ones and rearrange at the end.
That’s all I’ve got, I hope it helps!
PS:
Avoid photos with raindrops, severe insect damage, fungal infections, or other things that distort the appearance. That said, if that’s the only thing available, it’s better than nothing.
It should be mentioned that using the compare feature on the iNat website, you can see the first and the second photo.
This means that the second photo is nearly as important as the first one. Please choose a photo that shows a different aspect of the organism as the second photo. If the first photo shows a flower, then the second photo should show a general view of the plant, or a leaf or another diagnostic character. In butterflies, the first photo might show the upper side, make the second photo show the lower side (or the other sex in dimorphic species, or the caterpillar).
One more suggestion: If you know your plant’s genus has a homonym in another kingdom, look for an example that will most obviously present itself as “a plant” when appearing as a tiny thumbnail in the suggestion dropdown.
Ideally, someone should also choose a similarly clear “in this kingdom” example for the other homonym.
Good choice of primary and second picture, as seen in the compare view of the website.
First picture shows the round unsegmented trunk typical for the genus “Consolea” (A cactaceae often confused with Opuntia). The second picture shows details of the sculpture of the pads that are species specific, as well as flowers.
Bad choice of secondary picture, as seen in the compare view of the website.
First picture is not very good, as it shows a closeup of a flower but no leaves. At least the flower is typical for the genus. Second picture is a very bad choice as it does NOT show additional diagnostic characters like leaves, or total plant, but just another closeup of the flower.
Can you guess from these pictures that this is a vine? Probably not. A picture of a rare flower color or of isolated seeds makes sense as picture number 3 or 4. However not as the only additional character shown in picture number 2.
I agree that the first and second photos should show different aspects of the plant and should include parts that are important for identifying the species. Thanks for the examples.
I did want to underline how important # 8 is. Please do make sure that the photo really does still look clear,and good for ID, even at Thumbnail size!
I also wanted to emphasize that the first image should not be chosen on the basis of how “cool” it is, or what a great artistic photo it is in general, or whether it is your own particular favorite image for whatever reason.
Being a good clear image which helps show the way to the correct ID for the species even for a complete beginner, and even at a small size, is more important than how pretty or unusual the image is.
And I totally agree that in most cases a super close-up is inappropriate.
But in some cases the flower is the best way to ID the plant, so a closer image of the flower is not all bad.
I agree with a lot of what you said @jdjohnson. I would also use concepts of design to choose photos. Many users use their phones which makes the photos tiny, so I think it’s preferable to use photos where the colour of the background contrasts the subject, allowing the species to retain it’s shape even when small. I also think the subject should be centered in the image.
This feature request about attaching metadata to the photos in a plant taxon page could help a lot to identify observations, in particular observations showing only a flower, or only a fruit. (Add a “like” (over there) if you would like to get it!).
I can’t find it now, but I think for #1 there was a thread where someone mentioned that a lot of them could be fixed with a script. Basically if the picture for a species is from an inat observation and that inat observation is of a different species, this cannot be correct. And I definitely do run into cases like that from time to time.
Also, if the default picture is not from a RG inat observation (but flickr or non RG etc.), the script could probably flag those as well if an RG observation exists, in most cases that is a better choice.
I think that no script running in the background is needed. These triggers should be enough:
Trigger for checking (again) the consistency of an observation photo (already present in a taxon page) with the taxon: identification added/removed to an observation already flagged as source_of_taxon_photo=true
Trigger for updating the source_of_taxon_photo flag: photo added/removed to a taxon page.
Because a very small minority of observations would be flagged source_of_taxon_photo=true, the cost of the 1st trigger should be very low (hence no need for a update script running periodically in the background).
Because updating a taxon page is much less frequent than other actions (such as displaying/adding/identifying observations), the cost of the 2nd trigger should be very low as well.
